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 "mcast-snooping.h"
38 #include "meta-flow.h"
40 #include "ofp-print.h"
43 #include "ofproto/bond.h"
44 #include "ofproto/ofproto.h"
46 #include "poll-loop.h"
51 #include "socket-util.h"
53 #include "stream-ssl.h"
55 #include "system-stats.h"
60 #include "lib/vswitch-idl.h"
61 #include "xenserver.h"
63 #include "sflow_api.h"
64 #include "vlan-bitmap.h"
66 VLOG_DEFINE_THIS_MODULE(bridge);
68 COVERAGE_DEFINE(bridge_reconfigure);
71 /* These members are always valid.
73 * They are immutable: they never change between iface_create() and
75 struct list port_elem; /* Element in struct port's "ifaces" list. */
76 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
77 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
78 struct port *port; /* Containing port. */
79 char *name; /* Host network device name. */
80 struct netdev *netdev; /* Network device. */
81 ofp_port_t ofp_port; /* OpenFlow port number. */
84 /* These members are valid only within bridge_reconfigure(). */
85 const char *type; /* Usually same as cfg->type. */
86 const struct ovsrec_interface *cfg;
90 struct uuid uuid; /* UUID of this "mirror" record in database. */
91 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
92 struct bridge *bridge;
94 const struct ovsrec_mirror *cfg;
98 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
99 struct bridge *bridge;
102 const struct ovsrec_port *cfg;
104 /* An ordinary bridge port has 1 interface.
105 * A bridge port for bonding has at least 2 interfaces. */
106 struct list ifaces; /* List of "struct iface"s. */
110 struct hmap_node node; /* In 'all_bridges'. */
111 char *name; /* User-specified arbitrary name. */
112 char *type; /* Datapath type. */
113 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
114 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
115 const struct ovsrec_bridge *cfg;
117 /* OpenFlow switch processing. */
118 struct ofproto *ofproto; /* OpenFlow switch. */
121 struct hmap ports; /* "struct port"s indexed by name. */
122 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
123 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
125 /* Port mirroring. */
126 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
128 /* Used during reconfiguration. */
129 struct shash wanted_ports;
131 /* Synthetic local port if necessary. */
132 struct ovsrec_port synth_local_port;
133 struct ovsrec_interface synth_local_iface;
134 struct ovsrec_interface *synth_local_ifacep;
137 /* All bridges, indexed by name. */
138 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
140 /* OVSDB IDL used to obtain configuration. */
141 static struct ovsdb_idl *idl;
143 /* We want to complete daemonization, fully detaching from our parent process,
144 * only after we have completed our initial configuration, committed our state
145 * to the database, and received confirmation back from the database server
146 * that it applied the commit. This allows our parent process to know that,
147 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
148 * to have already been filled in. (It is only "very likely" rather than
149 * certain because there is always a slim possibility that the transaction will
150 * fail or that some other client has added new bridges, ports, etc. while
151 * ovs-vswitchd was configuring using an old configuration.)
153 * We only need to do this once for our initial configuration at startup, so
154 * 'initial_config_done' tracks whether we've already done it. While we are
155 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
156 * itself and is otherwise NULL. */
157 static bool initial_config_done;
158 static struct ovsdb_idl_txn *daemonize_txn;
160 /* Most recently processed IDL sequence number. */
161 static unsigned int idl_seqno;
163 /* Track changes to port connectivity. */
164 static uint64_t connectivity_seqno = LLONG_MIN;
166 /* Status update to database.
168 * Some information in the database must be kept as up-to-date as possible to
169 * allow controllers to respond rapidly to network outages. Those status are
170 * updated via the 'status_txn'.
172 * We use the global connectivity sequence number to detect the status change.
173 * Also, to prevent the status update from sending too much to the database,
174 * we check the return status of each update transaction and do not start new
175 * update if the previous transaction status is 'TXN_INCOMPLETE'.
177 * 'statux_txn' is NULL if there is no ongoing status update.
179 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
180 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
181 * which will cause the main thread wake up soon and retry the status update.
183 static struct ovsdb_idl_txn *status_txn;
184 static bool status_txn_try_again;
186 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
187 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
188 #define STATUS_CHECK_AGAIN_MSEC 100
190 /* Each time this timer expires, the bridge fetches interface and mirror
191 * statistics and pushes them into the database. */
192 static int stats_timer_interval;
193 static long long int stats_timer = LLONG_MIN;
195 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
196 * expensive. This can cause bridge_reconfigure() to take a long time during
197 * which no other work can be done. To deal with this problem, we limit port
198 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
199 * call to bridge_reconfigure(). If there is more work to do after the limit
200 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
201 * This allows the rest of the code to catch up on important things like
202 * forwarding packets. */
203 #define OFP_PORT_ACTION_WINDOW 10
205 static void add_del_bridges(const struct ovsrec_open_vswitch *);
206 static void bridge_run__(void);
207 static void bridge_create(const struct ovsrec_bridge *);
208 static void bridge_destroy(struct bridge *);
209 static struct bridge *bridge_lookup(const char *name);
210 static unixctl_cb_func bridge_unixctl_dump_flows;
211 static unixctl_cb_func bridge_unixctl_reconnect;
212 static size_t bridge_get_controllers(const struct bridge *br,
213 struct ovsrec_controller ***controllersp);
214 static void bridge_collect_wanted_ports(struct bridge *,
215 const unsigned long *splinter_vlans,
216 struct shash *wanted_ports);
217 static void bridge_delete_ofprotos(void);
218 static void bridge_delete_or_reconfigure_ports(struct bridge *);
219 static void bridge_del_ports(struct bridge *,
220 const struct shash *wanted_ports);
221 static void bridge_add_ports(struct bridge *,
222 const struct shash *wanted_ports);
224 static void bridge_configure_datapath_id(struct bridge *);
225 static void bridge_configure_netflow(struct bridge *);
226 static void bridge_configure_forward_bpdu(struct bridge *);
227 static void bridge_configure_mac_table(struct bridge *);
228 static void bridge_configure_mcast_snooping(struct bridge *);
229 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
230 static void bridge_configure_ipfix(struct bridge *);
231 static void bridge_configure_stp(struct bridge *);
232 static void bridge_configure_tables(struct bridge *);
233 static void bridge_configure_dp_desc(struct bridge *);
234 static void bridge_configure_remotes(struct bridge *,
235 const struct sockaddr_in *managers,
237 static void bridge_pick_local_hw_addr(struct bridge *,
238 uint8_t ea[ETH_ADDR_LEN],
239 struct iface **hw_addr_iface);
240 static uint64_t bridge_pick_datapath_id(struct bridge *,
241 const uint8_t bridge_ea[ETH_ADDR_LEN],
242 struct iface *hw_addr_iface);
243 static uint64_t dpid_from_hash(const void *, size_t nbytes);
244 static bool bridge_has_bond_fake_iface(const struct bridge *,
246 static bool port_is_bond_fake_iface(const struct port *);
248 static unixctl_cb_func qos_unixctl_show;
250 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
251 static void port_del_ifaces(struct port *);
252 static void port_destroy(struct port *);
253 static struct port *port_lookup(const struct bridge *, const char *name);
254 static void port_configure(struct port *);
255 static struct lacp_settings *port_configure_lacp(struct port *,
256 struct lacp_settings *);
257 static void port_configure_bond(struct port *, struct bond_settings *);
258 static bool port_is_synthetic(const struct port *);
260 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
261 static void run_system_stats(void);
263 static void bridge_configure_mirrors(struct bridge *);
264 static struct mirror *mirror_create(struct bridge *,
265 const struct ovsrec_mirror *);
266 static void mirror_destroy(struct mirror *);
267 static bool mirror_configure(struct mirror *);
268 static void mirror_refresh_stats(struct mirror *);
270 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
271 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
272 const struct ovsrec_port *);
273 static bool iface_is_internal(const struct ovsrec_interface *iface,
274 const struct ovsrec_bridge *br);
275 static const char *iface_get_type(const struct ovsrec_interface *,
276 const struct ovsrec_bridge *);
277 static void iface_destroy(struct iface *);
278 static void iface_destroy__(struct iface *);
279 static struct iface *iface_lookup(const struct bridge *, const char *name);
280 static struct iface *iface_find(const char *name);
281 static struct iface *iface_from_ofp_port(const struct bridge *,
282 ofp_port_t ofp_port);
283 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
284 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
285 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
286 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
287 static void iface_configure_cfm(struct iface *);
288 static void iface_refresh_cfm_stats(struct iface *);
289 static void iface_refresh_stats(struct iface *);
290 static void iface_refresh_netdev_status(struct iface *);
291 static void iface_refresh_ofproto_status(struct iface *);
292 static bool iface_is_synthetic(const struct iface *);
293 static ofp_port_t iface_get_requested_ofp_port(
294 const struct ovsrec_interface *);
295 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
297 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
299 * This is deprecated. It is only for compatibility with broken device drivers
300 * in old versions of Linux that do not properly support VLANs when VLAN
301 * devices are not used. When broken device drivers are no longer in
302 * widespread use, we will delete these interfaces. */
304 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
305 static bool vlan_splinters_enabled_anywhere;
307 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
308 static unsigned long int *collect_splinter_vlans(
309 const struct ovsrec_open_vswitch *);
310 static void configure_splinter_port(struct port *);
311 static void add_vlan_splinter_ports(struct bridge *,
312 const unsigned long int *splinter_vlans,
313 struct shash *ports);
316 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
318 struct shash iface_hints;
319 static bool initialized = false;
326 shash_init(&iface_hints);
329 for (i = 0; i < cfg->n_bridges; i++) {
330 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
333 for (j = 0; j < br_cfg->n_ports; j++) {
334 struct ovsrec_port *port_cfg = br_cfg->ports[j];
337 for (k = 0; k < port_cfg->n_interfaces; k++) {
338 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
339 struct iface_hint *iface_hint;
341 iface_hint = xmalloc(sizeof *iface_hint);
342 iface_hint->br_name = br_cfg->name;
343 iface_hint->br_type = br_cfg->datapath_type;
344 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
346 shash_add(&iface_hints, if_cfg->name, iface_hint);
352 ofproto_init(&iface_hints);
354 shash_destroy_free_data(&iface_hints);
358 /* Public functions. */
360 /* Initializes the bridge module, configuring it to obtain its configuration
361 * from an OVSDB server accessed over 'remote', which should be a string in a
362 * form acceptable to ovsdb_idl_create(). */
364 bridge_init(const char *remote)
366 /* Create connection to database. */
367 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
368 idl_seqno = ovsdb_idl_get_seqno(idl);
369 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
370 ovsdb_idl_verify_write_only(idl);
372 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
373 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
374 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
375 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
376 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
378 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
380 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
381 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
382 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
384 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
385 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
386 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
388 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
389 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
408 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
410 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
411 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
412 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
413 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
415 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
417 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
419 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
420 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
422 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
423 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
425 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
427 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
428 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
429 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
430 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
431 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
433 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
435 /* Register unixctl commands. */
436 unixctl_command_register("qos/show", "interface", 1, 1,
437 qos_unixctl_show, NULL);
438 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
439 bridge_unixctl_dump_flows, NULL);
440 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
441 bridge_unixctl_reconnect, NULL);
452 struct bridge *br, *next_br;
454 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
457 ovsdb_idl_destroy(idl);
460 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
461 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
462 * responsible for freeing '*managersp' (with free()).
464 * You may be asking yourself "why does ovs-vswitchd care?", because
465 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
466 * should not be and in fact is not directly involved in that. But
467 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
468 * it has to tell in-band control where the managers are to enable that.
469 * (Thus, only managers connected in-band are collected.)
472 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
473 struct sockaddr_in **managersp, size_t *n_managersp)
475 struct sockaddr_in *managers = NULL;
476 size_t n_managers = 0;
480 /* Collect all of the potential targets from the "targets" columns of the
481 * rows pointed to by "manager_options", excluding any that are
484 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
485 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
487 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
488 sset_find_and_delete(&targets, m->target);
490 sset_add(&targets, m->target);
494 /* Now extract the targets' IP addresses. */
495 if (!sset_is_empty(&targets)) {
498 managers = xmalloc(sset_count(&targets) * sizeof *managers);
499 SSET_FOR_EACH (target, &targets) {
501 struct sockaddr_storage ss;
502 struct sockaddr_in in;
505 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
507 && sa.ss.ss_family == AF_INET) {
508 managers[n_managers++] = sa.in;
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 /* Clear eventual previous errors */
613 ovsrec_interface_set_error(iface->cfg, NULL);
614 iface_configure_cfm(iface);
615 iface_configure_qos(iface, port->cfg->qos);
616 iface_set_mac(br, port, iface);
617 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
621 bridge_configure_mirrors(br);
622 bridge_configure_forward_bpdu(br);
623 bridge_configure_mac_table(br);
624 bridge_configure_mcast_snooping(br);
625 bridge_configure_remotes(br, managers, n_managers);
626 bridge_configure_netflow(br);
627 bridge_configure_sflow(br, &sflow_bridge_number);
628 bridge_configure_ipfix(br);
629 bridge_configure_stp(br);
630 bridge_configure_tables(br);
631 bridge_configure_dp_desc(br);
635 /* The ofproto-dpif provider does some final reconfiguration in its
636 * ->type_run() function. We have to call it before notifying the database
637 * client that reconfiguration is complete, otherwise there is a very
638 * narrow race window in which e.g. ofproto/trace will not recognize the
639 * new configuration (sometimes this causes unit test failures). */
643 /* Delete ofprotos which aren't configured or have the wrong type. Create
644 * ofprotos which don't exist but need to. */
646 bridge_delete_ofprotos(void)
653 /* Delete ofprotos with no bridge or with the wrong type. */
656 ofproto_enumerate_types(&types);
657 SSET_FOR_EACH (type, &types) {
660 ofproto_enumerate_names(type, &names);
661 SSET_FOR_EACH (name, &names) {
662 br = bridge_lookup(name);
663 if (!br || strcmp(type, br->type)) {
664 ofproto_delete(name, type);
668 sset_destroy(&names);
669 sset_destroy(&types);
673 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
675 if (*n >= *allocated) {
676 ports = x2nrealloc(ports, allocated, sizeof *ports);
678 ports[(*n)++] = port;
683 bridge_delete_or_reconfigure_ports(struct bridge *br)
685 struct ofproto_port ofproto_port;
686 struct ofproto_port_dump dump;
688 struct sset ofproto_ports;
689 struct port *port, *port_next;
691 /* List of "ofp_port"s to delete. We make a list instead of deleting them
692 * right away because ofproto implementations aren't necessarily able to
693 * iterate through a changing list of ports in an entirely robust way. */
700 sset_init(&ofproto_ports);
702 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
703 * that are not configured in the database. (This commonly happens when
704 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
706 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
707 * that have the wrong OpenFlow port number (and arrange to add them back
708 * with the correct OpenFlow port number). */
709 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
710 ofp_port_t requested_ofp_port;
713 sset_add(&ofproto_ports, ofproto_port.name);
715 iface = iface_lookup(br, ofproto_port.name);
717 /* No such iface is configured, so we should delete this
720 * As a corner case exception, keep the port if it's a bond fake
722 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
723 && !strcmp(ofproto_port.type, "internal")) {
729 if (strcmp(ofproto_port.type, iface->type)
730 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
731 /* The interface is the wrong type or can't be configured.
736 /* If the requested OpenFlow port for 'iface' changed, and it's not
737 * already the correct port, then we might want to temporarily delete
738 * this interface, so we can add it back again with the new OpenFlow
740 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
741 if (iface->ofp_port != OFPP_LOCAL &&
742 requested_ofp_port != OFPP_NONE &&
743 requested_ofp_port != iface->ofp_port) {
744 ofp_port_t victim_request;
745 struct iface *victim;
747 /* Check for an existing OpenFlow port currently occupying
748 * 'iface''s requested port number. If there isn't one, then
749 * delete this port. Otherwise we need to consider further. */
750 victim = iface_from_ofp_port(br, requested_ofp_port);
755 /* 'victim' is a port currently using 'iface''s requested port
756 * number. Unless 'victim' specifically requested that port
757 * number, too, then we can delete both 'iface' and 'victim'
758 * temporarily. (We'll add both of them back again later with new
759 * OpenFlow port numbers.)
761 * If 'victim' did request port number 'requested_ofp_port', just
762 * like 'iface', then that's a configuration inconsistency that we
763 * can't resolve. We might as well let it keep its current port
765 victim_request = iface_get_requested_ofp_port(victim->cfg);
766 if (victim_request != requested_ofp_port) {
767 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
768 iface_destroy(victim);
777 iface_destroy(iface);
778 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
780 for (i = 0; i < n; i++) {
781 ofproto_port_del(br->ofproto, del[i]);
785 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
786 * that we didn't see when we iterated through the datapath, i.e. ports
787 * that disappeared underneath use. This is an unusual situation, but it
788 * can happen in some cases:
790 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
791 * idea but could happen).
793 * - The port represented a device that disappeared, e.g. a tuntap
794 * device destroyed via "tunctl -d", a physical Ethernet device
795 * whose module was just unloaded via "rmmod", or a virtual NIC for a
796 * VM whose VM was just terminated. */
797 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
798 struct iface *iface, *iface_next;
800 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
801 if (!sset_contains(&ofproto_ports, iface->name)) {
802 iface_destroy__(iface);
806 if (list_is_empty(&port->ifaces)) {
810 sset_destroy(&ofproto_ports);
814 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
815 bool with_requested_port)
817 struct shash_node *port_node;
819 SHASH_FOR_EACH (port_node, wanted_ports) {
820 const struct ovsrec_port *port_cfg = port_node->data;
823 for (i = 0; i < port_cfg->n_interfaces; i++) {
824 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
825 ofp_port_t requested_ofp_port;
827 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
828 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
829 struct iface *iface = iface_lookup(br, iface_cfg->name);
832 iface_create(br, iface_cfg, port_cfg);
840 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
842 /* First add interfaces that request a particular port number. */
843 bridge_add_ports__(br, wanted_ports, true);
845 /* Then add interfaces that want automatic port number assignment.
846 * We add these afterward to avoid accidentally taking a specifically
847 * requested port number. */
848 bridge_add_ports__(br, wanted_ports, false);
852 port_configure(struct port *port)
854 const struct ovsrec_port *cfg = port->cfg;
855 struct bond_settings bond_settings;
856 struct lacp_settings lacp_settings;
857 struct ofproto_bundle_settings s;
860 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
861 configure_splinter_port(port);
870 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
871 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
872 s.slaves[s.n_slaves++] = iface->ofp_port;
877 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
881 /* Get VLAN trunks. */
884 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
888 if (cfg->vlan_mode) {
889 if (!strcmp(cfg->vlan_mode, "access")) {
890 s.vlan_mode = PORT_VLAN_ACCESS;
891 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
892 s.vlan_mode = PORT_VLAN_TRUNK;
893 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
894 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
895 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
896 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
898 /* This "can't happen" because ovsdb-server should prevent it. */
899 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
900 "back to trunk mode", port->name, cfg->vlan_mode);
901 s.vlan_mode = PORT_VLAN_TRUNK;
905 s.vlan_mode = PORT_VLAN_ACCESS;
907 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
911 s.vlan_mode = PORT_VLAN_TRUNK;
914 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
917 /* Get LACP settings. */
918 s.lacp = port_configure_lacp(port, &lacp_settings);
922 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
923 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
924 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
927 s.lacp_slaves = NULL;
930 /* Get bond settings. */
931 if (s.n_slaves > 1) {
932 s.bond = &bond_settings;
933 port_configure_bond(port, &bond_settings);
936 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
937 netdev_set_miimon_interval(iface->netdev, 0);
942 ofproto_bundle_register(port->bridge->ofproto, port, &s);
950 /* Pick local port hardware address and datapath ID for 'br'. */
952 bridge_configure_datapath_id(struct bridge *br)
954 uint8_t ea[ETH_ADDR_LEN];
956 struct iface *local_iface;
957 struct iface *hw_addr_iface;
960 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
961 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
963 int error = netdev_set_etheraddr(local_iface->netdev, ea);
965 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
966 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
967 "Ethernet address: %s",
968 br->name, ovs_strerror(error));
971 memcpy(br->ea, ea, ETH_ADDR_LEN);
973 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
974 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
975 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
976 ofproto_set_datapath_id(br->ofproto, dpid);
979 dpid_string = xasprintf("%016"PRIx64, dpid);
980 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
984 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
987 bridge_get_allowed_versions(struct bridge *br)
989 if (!br->cfg->n_protocols)
992 return ofputil_versions_from_strings(br->cfg->protocols,
993 br->cfg->n_protocols);
996 /* Set NetFlow configuration on 'br'. */
998 bridge_configure_netflow(struct bridge *br)
1000 struct ovsrec_netflow *cfg = br->cfg->netflow;
1001 struct netflow_options opts;
1004 ofproto_set_netflow(br->ofproto, NULL);
1008 memset(&opts, 0, sizeof opts);
1010 /* Get default NetFlow configuration from datapath.
1011 * Apply overrides from 'cfg'. */
1012 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1013 if (cfg->engine_type) {
1014 opts.engine_type = *cfg->engine_type;
1016 if (cfg->engine_id) {
1017 opts.engine_id = *cfg->engine_id;
1020 /* Configure active timeout interval. */
1021 opts.active_timeout = cfg->active_timeout;
1022 if (!opts.active_timeout) {
1023 opts.active_timeout = -1;
1024 } else if (opts.active_timeout < 0) {
1025 VLOG_WARN("bridge %s: active timeout interval set to negative "
1026 "value, using default instead (%d seconds)", br->name,
1027 NF_ACTIVE_TIMEOUT_DEFAULT);
1028 opts.active_timeout = -1;
1031 /* Add engine ID to interface number to disambiguate bridgs? */
1032 opts.add_id_to_iface = cfg->add_id_to_interface;
1033 if (opts.add_id_to_iface) {
1034 if (opts.engine_id > 0x7f) {
1035 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1036 "another vswitch, choose an engine id less than 128",
1039 if (hmap_count(&br->ports) > 508) {
1040 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1041 "another port when more than 508 ports are used",
1047 sset_init(&opts.collectors);
1048 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1051 if (ofproto_set_netflow(br->ofproto, &opts)) {
1052 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1054 sset_destroy(&opts.collectors);
1057 /* Set sFlow configuration on 'br'. */
1059 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1061 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1062 struct ovsrec_controller **controllers;
1063 struct ofproto_sflow_options oso;
1064 size_t n_controllers;
1068 ofproto_set_sflow(br->ofproto, NULL);
1072 memset(&oso, 0, sizeof oso);
1074 sset_init(&oso.targets);
1075 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1077 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1078 if (cfg->sampling) {
1079 oso.sampling_rate = *cfg->sampling;
1082 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1084 oso.polling_interval = *cfg->polling;
1087 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1089 oso.header_len = *cfg->header;
1092 oso.sub_id = (*sflow_bridge_number)++;
1093 oso.agent_device = cfg->agent;
1095 oso.control_ip = NULL;
1096 n_controllers = bridge_get_controllers(br, &controllers);
1097 for (i = 0; i < n_controllers; i++) {
1098 if (controllers[i]->local_ip) {
1099 oso.control_ip = controllers[i]->local_ip;
1103 ofproto_set_sflow(br->ofproto, &oso);
1105 sset_destroy(&oso.targets);
1108 /* Returns whether a IPFIX row is valid. */
1110 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1112 return ipfix && ipfix->n_targets > 0;
1115 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1117 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1118 const struct bridge *br)
1120 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1123 /* Set IPFIX configuration on 'br'. */
1125 bridge_configure_ipfix(struct bridge *br)
1127 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1128 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1129 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1130 struct ofproto_ipfix_bridge_exporter_options be_opts;
1131 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1132 size_t n_fe_opts = 0;
1134 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1135 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1140 if (!valid_be_cfg && n_fe_opts == 0) {
1141 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1146 memset(&be_opts, 0, sizeof be_opts);
1148 sset_init(&be_opts.targets);
1149 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1151 if (be_cfg->sampling) {
1152 be_opts.sampling_rate = *be_cfg->sampling;
1154 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1156 if (be_cfg->obs_domain_id) {
1157 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1159 if (be_cfg->obs_point_id) {
1160 be_opts.obs_point_id = *be_cfg->obs_point_id;
1162 if (be_cfg->cache_active_timeout) {
1163 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1165 if (be_cfg->cache_max_flows) {
1166 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1170 if (n_fe_opts > 0) {
1171 struct ofproto_ipfix_flow_exporter_options *opts;
1172 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1174 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1175 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1176 opts->collector_set_id = fe_cfg->id;
1177 sset_init(&opts->targets);
1178 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1179 fe_cfg->ipfix->n_targets);
1180 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1181 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1182 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1183 ? *fe_cfg->ipfix->cache_max_flows : 0;
1189 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1193 sset_destroy(&be_opts.targets);
1196 if (n_fe_opts > 0) {
1197 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1199 for (i = 0; i < n_fe_opts; i++) {
1200 sset_destroy(&opts->targets);
1208 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1209 struct ofproto_port_stp_settings *port_s,
1210 int *port_num_counter, unsigned long *port_num_bitmap)
1212 const char *config_str;
1213 struct iface *iface;
1215 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1216 port_s->enable = false;
1219 port_s->enable = true;
1222 /* STP over bonds is not supported. */
1223 if (!list_is_singleton(&port->ifaces)) {
1224 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1226 port_s->enable = false;
1230 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1232 /* Internal ports shouldn't participate in spanning tree, so
1234 if (!strcmp(iface->type, "internal")) {
1235 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1236 port_s->enable = false;
1240 /* STP on mirror output ports is not supported. */
1241 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1242 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1243 port_s->enable = false;
1247 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1249 unsigned long int port_num = strtoul(config_str, NULL, 0);
1250 int port_idx = port_num - 1;
1252 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1253 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1254 port_s->enable = false;
1258 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1259 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1260 port->name, port_num);
1261 port_s->enable = false;
1264 bitmap_set1(port_num_bitmap, port_idx);
1265 port_s->port_num = port_idx;
1267 if (*port_num_counter >= STP_MAX_PORTS) {
1268 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1269 port_s->enable = false;
1273 port_s->port_num = (*port_num_counter)++;
1276 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1278 port_s->path_cost = strtoul(config_str, NULL, 10);
1280 enum netdev_features current;
1283 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1284 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1285 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1288 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1290 port_s->priority = strtoul(config_str, NULL, 0);
1292 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1296 /* Set spanning tree configuration on 'br'. */
1298 bridge_configure_stp(struct bridge *br)
1300 if (!br->cfg->stp_enable) {
1301 ofproto_set_stp(br->ofproto, NULL);
1303 struct ofproto_stp_settings br_s;
1304 const char *config_str;
1306 int port_num_counter;
1307 unsigned long *port_num_bitmap;
1309 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1311 uint8_t ea[ETH_ADDR_LEN];
1313 if (eth_addr_from_string(config_str, ea)) {
1314 br_s.system_id = eth_addr_to_uint64(ea);
1316 br_s.system_id = eth_addr_to_uint64(br->ea);
1317 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1318 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1321 br_s.system_id = eth_addr_to_uint64(br->ea);
1324 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1326 br_s.priority = strtoul(config_str, NULL, 0);
1328 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1331 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1333 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1335 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1338 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1340 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1342 br_s.max_age = STP_DEFAULT_MAX_AGE;
1345 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1347 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1349 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1352 /* Configure STP on the bridge. */
1353 if (ofproto_set_stp(br->ofproto, &br_s)) {
1354 VLOG_ERR("bridge %s: could not enable STP", br->name);
1358 /* Users must either set the port number with the "stp-port-num"
1359 * configuration on all ports or none. If manual configuration
1360 * is not done, then we allocate them sequentially. */
1361 port_num_counter = 0;
1362 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1363 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1364 struct ofproto_port_stp_settings port_s;
1365 struct iface *iface;
1367 port_configure_stp(br->ofproto, port, &port_s,
1368 &port_num_counter, port_num_bitmap);
1370 /* As bonds are not supported, just apply configuration to
1371 * all interfaces. */
1372 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1373 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1375 VLOG_ERR("port %s: could not enable STP", port->name);
1381 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1382 && port_num_counter) {
1383 VLOG_ERR("bridge %s: must manually configure all STP port "
1384 "IDs or none, disabling", br->name);
1385 ofproto_set_stp(br->ofproto, NULL);
1387 bitmap_free(port_num_bitmap);
1392 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1394 const struct port *port = port_lookup(br, name);
1395 return port && port_is_bond_fake_iface(port);
1399 port_is_bond_fake_iface(const struct port *port)
1401 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1405 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1407 struct bridge *br, *next;
1408 struct shash new_br;
1411 /* Collect new bridges' names and types. */
1412 shash_init(&new_br);
1413 for (i = 0; i < cfg->n_bridges; i++) {
1414 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1415 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1417 if (strchr(br_cfg->name, '/')) {
1418 /* Prevent remote ovsdb-server users from accessing arbitrary
1419 * directories, e.g. consider a bridge named "../../../etc/". */
1420 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1422 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1423 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1427 /* Get rid of deleted bridges or those whose types have changed.
1428 * Update 'cfg' of bridges that still exist. */
1429 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1430 br->cfg = shash_find_data(&new_br, br->name);
1431 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1432 br->cfg->datapath_type))) {
1437 /* Add new bridges. */
1438 for (i = 0; i < cfg->n_bridges; i++) {
1439 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1440 struct bridge *br = bridge_lookup(br_cfg->name);
1442 bridge_create(br_cfg);
1446 shash_destroy(&new_br);
1449 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1450 * Returns 0 if successful, otherwise a positive errno value. */
1452 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1453 struct netdev *netdev, char **errp)
1455 return netdev_set_config(netdev, &iface_cfg->options, errp);
1458 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1459 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1461 * If successful, returns 0 and stores the network device in '*netdevp'. On
1462 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1464 iface_do_create(const struct bridge *br,
1465 const struct ovsrec_interface *iface_cfg,
1466 const struct ovsrec_port *port_cfg,
1467 ofp_port_t *ofp_portp, struct netdev **netdevp,
1470 struct netdev *netdev = NULL;
1473 if (netdev_is_reserved_name(iface_cfg->name)) {
1474 VLOG_WARN("could not create interface %s, name is reserved",
1480 error = netdev_open(iface_cfg->name,
1481 iface_get_type(iface_cfg, br->cfg), &netdev);
1483 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1484 iface_cfg->name, ovs_strerror(error));
1488 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1493 *ofp_portp = iface_pick_ofport(iface_cfg);
1494 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1499 VLOG_INFO("bridge %s: added interface %s on port %d",
1500 br->name, iface_cfg->name, *ofp_portp);
1502 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1503 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1511 netdev_close(netdev);
1515 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1516 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1517 * automatically allocated for the iface. Takes ownership of and
1518 * deallocates 'if_cfg'.
1520 * Return true if an iface is successfully created, false otherwise. */
1522 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1523 const struct ovsrec_port *port_cfg)
1525 struct netdev *netdev;
1526 struct iface *iface;
1527 ofp_port_t ofp_port;
1532 /* Do the bits that can fail up front. */
1533 ovs_assert(!iface_lookup(br, iface_cfg->name));
1534 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1536 iface_clear_db_record(iface_cfg, errp);
1541 /* Get or create the port structure. */
1542 port = port_lookup(br, port_cfg->name);
1544 port = port_create(br, port_cfg);
1547 /* Create the iface structure. */
1548 iface = xzalloc(sizeof *iface);
1549 list_push_back(&port->ifaces, &iface->port_elem);
1550 hmap_insert(&br->iface_by_name, &iface->name_node,
1551 hash_string(iface_cfg->name, 0));
1553 iface->name = xstrdup(iface_cfg->name);
1554 iface->ofp_port = ofp_port;
1555 iface->netdev = netdev;
1556 iface->type = iface_get_type(iface_cfg, br->cfg);
1557 iface->cfg = iface_cfg;
1558 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1559 hash_ofp_port(ofp_port));
1561 /* Populate initial status in database. */
1562 iface_refresh_stats(iface);
1563 iface_refresh_netdev_status(iface);
1565 /* Add bond fake iface if necessary. */
1566 if (port_is_bond_fake_iface(port)) {
1567 struct ofproto_port ofproto_port;
1569 if (ofproto_port_query_by_name(br->ofproto, port->name,
1571 struct netdev *netdev;
1574 error = netdev_open(port->name, "internal", &netdev);
1576 ofp_port_t fake_ofp_port = OFPP_NONE;
1577 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1578 netdev_close(netdev);
1580 VLOG_WARN("could not open network device %s (%s)",
1581 port->name, ovs_strerror(error));
1584 /* Already exists, nothing to do. */
1585 ofproto_port_destroy(&ofproto_port);
1592 /* Set forward BPDU option. */
1594 bridge_configure_forward_bpdu(struct bridge *br)
1596 ofproto_set_forward_bpdu(br->ofproto,
1597 smap_get_bool(&br->cfg->other_config,
1602 /* Set MAC learning table configuration for 'br'. */
1604 bridge_configure_mac_table(struct bridge *br)
1606 const char *idle_time_str;
1609 const char *mac_table_size_str;
1612 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1613 idle_time = (idle_time_str && atoi(idle_time_str)
1614 ? atoi(idle_time_str)
1615 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1617 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1618 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1619 ? atoi(mac_table_size_str)
1622 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1625 /* Set multicast snooping table configuration for 'br'. */
1627 bridge_configure_mcast_snooping(struct bridge *br)
1629 if (!br->cfg->mcast_snooping_enable) {
1630 ofproto_set_mcast_snooping(br->ofproto, NULL);
1633 struct ofproto_mcast_snooping_settings br_s;
1634 const char *idle_time_str;
1635 const char *max_entries_str;
1637 idle_time_str = smap_get(&br->cfg->other_config,
1638 "mcast-snooping-aging-time");
1639 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1640 ? atoi(idle_time_str)
1641 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1643 max_entries_str = smap_get(&br->cfg->other_config,
1644 "mcast-snooping-table-size");
1645 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1646 ? atoi(max_entries_str)
1647 : MCAST_DEFAULT_MAX_ENTRIES);
1649 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1650 "mcast-snooping-disable-flood-unregistered",
1653 /* Configure multicast snooping on the bridge */
1654 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1655 VLOG_ERR("bridge %s: could not enable multicast snooping",
1660 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1661 bool flood = smap_get_bool(&port->cfg->other_config,
1662 "mcast-snooping-flood", false);
1663 if (ofproto_port_set_mcast_snooping(br->ofproto, port, flood)) {
1664 VLOG_ERR("port %s: could not configure mcast snooping",
1672 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1673 const struct port *fake_br, struct iface **hw_addr_iface)
1675 struct hmapx mirror_output_ports;
1677 bool found_addr = false;
1681 /* Mirror output ports don't participate in picking the local hardware
1682 * address. ofproto can't help us find out whether a given port is a
1683 * mirror output because we haven't configured mirrors yet, so we need to
1684 * accumulate them ourselves. */
1685 hmapx_init(&mirror_output_ports);
1686 for (i = 0; i < br->cfg->n_mirrors; i++) {
1687 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1688 if (m->output_port) {
1689 hmapx_add(&mirror_output_ports, m->output_port);
1693 /* Otherwise choose the minimum non-local MAC address among all of the
1695 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1696 uint8_t iface_ea[ETH_ADDR_LEN];
1697 struct iface *candidate;
1698 struct iface *iface;
1700 /* Mirror output ports don't participate. */
1701 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1705 /* Choose the MAC address to represent the port. */
1707 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1708 /* Find the interface with this Ethernet address (if any) so that
1709 * we can provide the correct devname to the caller. */
1710 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1711 uint8_t candidate_ea[ETH_ADDR_LEN];
1712 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1713 && eth_addr_equals(iface_ea, candidate_ea)) {
1718 /* Choose the interface whose MAC address will represent the port.
1719 * The Linux kernel bonding code always chooses the MAC address of
1720 * the first slave added to a bond, and the Fedora networking
1721 * scripts always add slaves to a bond in alphabetical order, so
1722 * for compatibility we choose the interface with the name that is
1723 * first in alphabetical order. */
1724 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1725 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1730 /* The local port doesn't count (since we're trying to choose its
1731 * MAC address anyway). */
1732 if (iface->ofp_port == OFPP_LOCAL) {
1736 /* For fake bridges we only choose from ports with the same tag */
1737 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1738 if (!port->cfg->tag) {
1741 if (*port->cfg->tag != *fake_br->cfg->tag) {
1747 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1753 /* Compare against our current choice. */
1754 if (!eth_addr_is_multicast(iface_ea) &&
1755 !eth_addr_is_local(iface_ea) &&
1756 !eth_addr_is_reserved(iface_ea) &&
1757 !eth_addr_is_zero(iface_ea) &&
1758 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1760 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1761 *hw_addr_iface = iface;
1767 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1768 *hw_addr_iface = NULL;
1771 hmapx_destroy(&mirror_output_ports);
1775 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1776 struct iface **hw_addr_iface)
1779 *hw_addr_iface = NULL;
1781 /* Did the user request a particular MAC? */
1782 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1783 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1784 if (eth_addr_is_multicast(ea)) {
1785 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1786 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1787 } else if (eth_addr_is_zero(ea)) {
1788 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1794 /* Find a local hw address */
1795 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1798 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1799 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1800 * an interface on 'br', then that interface must be passed in as
1801 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1802 * 'hw_addr_iface' must be passed in as a null pointer. */
1804 bridge_pick_datapath_id(struct bridge *br,
1805 const uint8_t bridge_ea[ETH_ADDR_LEN],
1806 struct iface *hw_addr_iface)
1809 * The procedure for choosing a bridge MAC address will, in the most
1810 * ordinary case, also choose a unique MAC that we can use as a datapath
1811 * ID. In some special cases, though, multiple bridges will end up with
1812 * the same MAC address. This is OK for the bridges, but it will confuse
1813 * the OpenFlow controller, because each datapath needs a unique datapath
1816 * Datapath IDs must be unique. It is also very desirable that they be
1817 * stable from one run to the next, so that policy set on a datapath
1820 const char *datapath_id;
1823 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1824 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1828 if (!hw_addr_iface) {
1830 * A purely internal bridge, that is, one that has no non-virtual
1831 * network devices on it at all, is difficult because it has no
1832 * natural unique identifier at all.
1834 * When the host is a XenServer, we handle this case by hashing the
1835 * host's UUID with the name of the bridge. Names of bridges are
1836 * persistent across XenServer reboots, although they can be reused if
1837 * an internal network is destroyed and then a new one is later
1838 * created, so this is fairly effective.
1840 * When the host is not a XenServer, we punt by using a random MAC
1841 * address on each run.
1843 const char *host_uuid = xenserver_get_host_uuid();
1845 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1846 dpid = dpid_from_hash(combined, strlen(combined));
1852 return eth_addr_to_uint64(bridge_ea);
1856 dpid_from_hash(const void *data, size_t n)
1858 uint8_t hash[SHA1_DIGEST_SIZE];
1860 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1861 sha1_bytes(data, n, hash);
1862 eth_addr_mark_random(hash);
1863 return eth_addr_to_uint64(hash);
1867 iface_refresh_netdev_status(struct iface *iface)
1871 enum netdev_features current;
1872 enum netdev_flags flags;
1873 const char *link_state;
1874 uint8_t mac[ETH_ADDR_LEN];
1875 int64_t bps, mtu_64, ifindex64, link_resets;
1878 if (iface_is_synthetic(iface)) {
1882 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
1883 && !status_txn_try_again) {
1887 iface->change_seq = netdev_get_change_seq(iface->netdev);
1891 if (!netdev_get_status(iface->netdev, &smap)) {
1892 ovsrec_interface_set_status(iface->cfg, &smap);
1894 ovsrec_interface_set_status(iface->cfg, NULL);
1897 smap_destroy(&smap);
1899 error = netdev_get_flags(iface->netdev, &flags);
1901 const char *state = flags & NETDEV_UP ? "up" : "down";
1903 ovsrec_interface_set_admin_state(iface->cfg, state);
1905 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1908 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1909 ovsrec_interface_set_link_state(iface->cfg, link_state);
1911 link_resets = netdev_get_carrier_resets(iface->netdev);
1912 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1914 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1915 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1917 ovsrec_interface_set_duplex(iface->cfg,
1918 netdev_features_is_full_duplex(current)
1920 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1922 ovsrec_interface_set_duplex(iface->cfg, NULL);
1923 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1926 error = netdev_get_mtu(iface->netdev, &mtu);
1929 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1931 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1934 error = netdev_get_etheraddr(iface->netdev, mac);
1936 char mac_string[32];
1938 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1939 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1941 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1944 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1945 * if there is no valid ifindex number. */
1946 ifindex64 = netdev_get_ifindex(iface->netdev);
1947 if (ifindex64 < 0) {
1950 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1954 iface_refresh_ofproto_status(struct iface *iface)
1958 if (iface_is_synthetic(iface)) {
1962 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1966 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1968 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1971 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
1973 || status_txn_try_again) {
1974 iface_refresh_cfm_stats(iface);
1977 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
1979 || status_txn_try_again) {
1983 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1984 iface->ofp_port, &smap);
1985 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1986 smap_destroy(&smap);
1990 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1993 iface_refresh_cfm_stats(struct iface *iface)
1995 const struct ovsrec_interface *cfg = iface->cfg;
1996 struct cfm_status status;
1999 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2000 iface->ofp_port, &status);
2002 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2003 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2004 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2005 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2006 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2007 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2009 const char *reasons[CFM_FAULT_N_REASONS];
2010 int64_t cfm_health = status.health;
2011 int64_t cfm_flap_count = status.flap_count;
2012 bool faulted = status.faults != 0;
2015 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2018 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2019 int reason = 1 << i;
2020 if (status.faults & reason) {
2021 reasons[j++] = cfm_fault_reason_to_str(reason);
2024 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
2026 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2028 if (status.remote_opstate >= 0) {
2029 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2030 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2032 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2035 ovsrec_interface_set_cfm_remote_mpids(cfg,
2036 (const int64_t *)status.rmps,
2038 if (cfm_health >= 0) {
2039 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2041 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2049 iface_refresh_stats(struct iface *iface)
2051 #define IFACE_STATS \
2052 IFACE_STAT(rx_packets, "rx_packets") \
2053 IFACE_STAT(tx_packets, "tx_packets") \
2054 IFACE_STAT(rx_bytes, "rx_bytes") \
2055 IFACE_STAT(tx_bytes, "tx_bytes") \
2056 IFACE_STAT(rx_dropped, "rx_dropped") \
2057 IFACE_STAT(tx_dropped, "tx_dropped") \
2058 IFACE_STAT(rx_errors, "rx_errors") \
2059 IFACE_STAT(tx_errors, "tx_errors") \
2060 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2061 IFACE_STAT(rx_over_errors, "rx_over_err") \
2062 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2063 IFACE_STAT(collisions, "collisions")
2065 #define IFACE_STAT(MEMBER, NAME) + 1
2066 enum { N_IFACE_STATS = IFACE_STATS };
2068 int64_t values[N_IFACE_STATS];
2069 char *keys[N_IFACE_STATS];
2072 struct netdev_stats stats;
2074 if (iface_is_synthetic(iface)) {
2078 /* Intentionally ignore return value, since errors will set 'stats' to
2079 * all-1s, and we will deal with that correctly below. */
2080 netdev_get_stats(iface->netdev, &stats);
2082 /* Copy statistics into keys[] and values[]. */
2084 #define IFACE_STAT(MEMBER, NAME) \
2085 if (stats.MEMBER != UINT64_MAX) { \
2087 values[n] = stats.MEMBER; \
2092 ovs_assert(n <= N_IFACE_STATS);
2094 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2099 br_refresh_stp_status(struct bridge *br)
2101 struct smap smap = SMAP_INITIALIZER(&smap);
2102 struct ofproto *ofproto = br->ofproto;
2103 struct ofproto_stp_status status;
2105 if (ofproto_get_stp_status(ofproto, &status)) {
2109 if (!status.enabled) {
2110 ovsrec_bridge_set_status(br->cfg, NULL);
2114 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2115 STP_ID_ARGS(status.bridge_id));
2116 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2117 STP_ID_ARGS(status.designated_root));
2118 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2120 ovsrec_bridge_set_status(br->cfg, &smap);
2121 smap_destroy(&smap);
2125 port_refresh_stp_status(struct port *port)
2127 struct ofproto *ofproto = port->bridge->ofproto;
2128 struct iface *iface;
2129 struct ofproto_port_stp_status status;
2132 if (port_is_synthetic(port)) {
2136 /* STP doesn't currently support bonds. */
2137 if (!list_is_singleton(&port->ifaces)) {
2138 ovsrec_port_set_status(port->cfg, NULL);
2142 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2143 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2147 if (!status.enabled) {
2148 ovsrec_port_set_status(port->cfg, NULL);
2152 /* Set Status column. */
2154 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2155 smap_add(&smap, "stp_state", stp_state_name(status.state));
2156 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2157 smap_add(&smap, "stp_role", stp_role_name(status.role));
2158 ovsrec_port_set_status(port->cfg, &smap);
2159 smap_destroy(&smap);
2163 port_refresh_stp_stats(struct port *port)
2165 struct ofproto *ofproto = port->bridge->ofproto;
2166 struct iface *iface;
2167 struct ofproto_port_stp_stats stats;
2169 int64_t int_values[3];
2171 if (port_is_synthetic(port)) {
2175 /* STP doesn't currently support bonds. */
2176 if (!list_is_singleton(&port->ifaces)) {
2180 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2181 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2185 if (!stats.enabled) {
2186 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2190 /* Set Statistics column. */
2191 keys[0] = "stp_tx_count";
2192 int_values[0] = stats.tx_count;
2193 keys[1] = "stp_rx_count";
2194 int_values[1] = stats.rx_count;
2195 keys[2] = "stp_error_count";
2196 int_values[2] = stats.error_count;
2198 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2199 ARRAY_SIZE(int_values));
2203 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2205 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2209 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2211 bool enable = enable_system_stats(cfg);
2213 system_stats_enable(enable);
2215 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2220 run_system_stats(void)
2222 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2225 stats = system_stats_run();
2227 struct ovsdb_idl_txn *txn;
2228 struct ovsdb_datum datum;
2230 txn = ovsdb_idl_txn_create(idl);
2231 ovsdb_datum_from_smap(&datum, stats);
2232 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2234 ovsdb_idl_txn_commit(txn);
2235 ovsdb_idl_txn_destroy(txn);
2242 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2245 case OFPCR12_ROLE_EQUAL:
2247 case OFPCR12_ROLE_MASTER:
2249 case OFPCR12_ROLE_SLAVE:
2251 case OFPCR12_ROLE_NOCHANGE:
2253 return "*** INVALID ROLE ***";
2258 refresh_controller_status(void)
2262 const struct ovsrec_controller *cfg;
2266 /* Accumulate status for controllers on all bridges. */
2267 HMAP_FOR_EACH (br, node, &all_bridges) {
2268 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2271 /* Update each controller in the database with current status. */
2272 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2273 struct ofproto_controller_info *cinfo =
2274 shash_find_data(&info, cfg->target);
2277 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2278 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2280 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2282 ovsrec_controller_set_is_connected(cfg, false);
2283 ovsrec_controller_set_role(cfg, NULL);
2284 ovsrec_controller_set_status(cfg, NULL);
2288 ofproto_free_ofproto_controller_info(&info);
2298 /* Let each datapath type do the work that it needs to do. */
2300 ofproto_enumerate_types(&types);
2301 SSET_FOR_EACH (type, &types) {
2302 ofproto_type_run(type);
2304 sset_destroy(&types);
2306 /* Let each bridge do the work that it needs to do. */
2307 HMAP_FOR_EACH (br, node, &all_bridges) {
2308 ofproto_run(br->ofproto);
2315 static struct ovsrec_open_vswitch null_cfg;
2316 const struct ovsrec_open_vswitch *cfg;
2318 bool vlan_splinters_changed;
2322 ovsrec_open_vswitch_init(&null_cfg);
2326 if (ovsdb_idl_is_lock_contended(idl)) {
2327 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2328 struct bridge *br, *next_br;
2330 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2331 "disabling this process (pid %ld) until it goes away",
2332 (long int) getpid());
2334 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2337 /* Since we will not be running system_stats_run() in this process
2338 * with the current situation of multiple ovs-vswitchd daemons,
2339 * disable system stats collection. */
2340 system_stats_enable(false);
2342 } else if (!ovsdb_idl_has_lock(idl)) {
2345 cfg = ovsrec_open_vswitch_first(idl);
2347 /* Initialize the ofproto library. This only needs to run once, but
2348 * it must be done after the configuration is set. If the
2349 * initialization has already occurred, bridge_init_ofproto()
2350 * returns immediately. */
2351 bridge_init_ofproto(cfg);
2353 /* Once the value of flow-restore-wait is false, we no longer should
2354 * check its value from the database. */
2355 if (cfg && ofproto_get_flow_restore_wait()) {
2356 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2357 "flow-restore-wait", false));
2362 /* Re-configure SSL. We do this on every trip through the main loop,
2363 * instead of just when the database changes, because the contents of the
2364 * key and certificate files can change without the database changing.
2366 * We do this before bridge_reconfigure() because that function might
2367 * initiate SSL connections and thus requires SSL to be configured. */
2368 if (cfg && cfg->ssl) {
2369 const struct ovsrec_ssl *ssl = cfg->ssl;
2371 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2372 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2375 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2376 * usage has changed. */
2377 vlan_splinters_changed = false;
2378 if (vlan_splinters_enabled_anywhere) {
2379 HMAP_FOR_EACH (br, node, &all_bridges) {
2380 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2381 vlan_splinters_changed = true;
2387 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2388 struct ovsdb_idl_txn *txn;
2390 idl_seqno = ovsdb_idl_get_seqno(idl);
2391 txn = ovsdb_idl_txn_create(idl);
2392 bridge_reconfigure(cfg ? cfg : &null_cfg);
2395 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2398 /* If we are completing our initial configuration for this run
2399 * of ovs-vswitchd, then keep the transaction around to monitor
2400 * it for completion. */
2401 if (initial_config_done) {
2402 /* Always sets the 'status_txn_try_again' to check again,
2403 * in case that this transaction fails. */
2404 status_txn_try_again = true;
2405 ovsdb_idl_txn_commit(txn);
2406 ovsdb_idl_txn_destroy(txn);
2408 initial_config_done = true;
2409 daemonize_txn = txn;
2413 if (daemonize_txn) {
2414 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2415 if (status != TXN_INCOMPLETE) {
2416 ovsdb_idl_txn_destroy(daemonize_txn);
2417 daemonize_txn = NULL;
2419 /* ovs-vswitchd has completed initialization, so allow the
2420 * process that forked us to exit successfully. */
2421 daemonize_complete();
2423 vlog_enable_async();
2425 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2429 /* Statistics update interval should always be greater than or equal to
2432 stats_interval = MAX(smap_get_int(&cfg->other_config,
2433 "stats-update-interval",
2436 stats_interval = 5000;
2438 if (stats_timer_interval != stats_interval) {
2439 stats_timer_interval = stats_interval;
2440 stats_timer = LLONG_MIN;
2443 /* Refresh interface and mirror stats if necessary. */
2444 if (time_msec() >= stats_timer) {
2446 struct ovsdb_idl_txn *txn;
2448 txn = ovsdb_idl_txn_create(idl);
2449 HMAP_FOR_EACH (br, node, &all_bridges) {
2453 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2454 struct iface *iface;
2456 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2457 iface_refresh_stats(iface);
2460 port_refresh_stp_stats(port);
2463 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2464 mirror_refresh_stats(m);
2468 refresh_controller_status();
2469 ovsdb_idl_txn_commit(txn);
2470 ovsdb_idl_txn_destroy(txn); /* XXX */
2473 stats_timer = time_msec() + stats_timer_interval;
2479 /* Check the need to update status. */
2480 seq = seq_read(connectivity_seq_get());
2481 if (seq != connectivity_seqno || status_txn_try_again) {
2482 connectivity_seqno = seq;
2483 status_txn = ovsdb_idl_txn_create(idl);
2484 HMAP_FOR_EACH (br, node, &all_bridges) {
2487 br_refresh_stp_status(br);
2488 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2489 struct iface *iface;
2491 port_refresh_stp_status(port);
2492 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2493 iface_refresh_netdev_status(iface);
2494 iface_refresh_ofproto_status(iface);
2502 enum ovsdb_idl_txn_status status;
2504 status = ovsdb_idl_txn_commit(status_txn);
2505 /* Do not destroy "status_txn" if the transaction is
2506 * "TXN_INCOMPLETE". */
2507 if (status != TXN_INCOMPLETE) {
2508 ovsdb_idl_txn_destroy(status_txn);
2511 /* Sets the 'status_txn_try_again' if the transaction fails. */
2512 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2513 status_txn_try_again = false;
2515 status_txn_try_again = true;
2529 ovsdb_idl_wait(idl);
2530 if (daemonize_txn) {
2531 ovsdb_idl_txn_wait(daemonize_txn);
2535 ofproto_enumerate_types(&types);
2536 SSET_FOR_EACH (type, &types) {
2537 ofproto_type_wait(type);
2539 sset_destroy(&types);
2541 if (!hmap_is_empty(&all_bridges)) {
2544 HMAP_FOR_EACH (br, node, &all_bridges) {
2545 ofproto_wait(br->ofproto);
2548 poll_timer_wait_until(stats_timer);
2551 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2552 * transaction to complete. If the status update to database needs to be
2553 * run again (transaction fails), registers a timeout in
2554 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2555 * sequence number. */
2557 ovsdb_idl_txn_wait(status_txn);
2558 } else if (status_txn_try_again) {
2559 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2561 seq_wait(connectivity_seq_get(), connectivity_seqno);
2564 system_stats_wait();
2567 /* Adds some memory usage statistics for bridges into 'usage', for use with
2568 * memory_report(). */
2570 bridge_get_memory_usage(struct simap *usage)
2577 ofproto_enumerate_types(&types);
2578 SSET_FOR_EACH (type, &types) {
2579 ofproto_type_get_memory_usage(type, usage);
2581 sset_destroy(&types);
2583 HMAP_FOR_EACH (br, node, &all_bridges) {
2584 ofproto_get_memory_usage(br->ofproto, usage);
2588 /* QoS unixctl user interface functions. */
2590 struct qos_unixctl_show_cbdata {
2592 struct iface *iface;
2596 qos_unixctl_show_queue(unsigned int queue_id,
2597 const struct smap *details,
2598 struct iface *iface,
2601 struct netdev_queue_stats stats;
2602 struct smap_node *node;
2605 ds_put_cstr(ds, "\n");
2607 ds_put_format(ds, "Queue %u:\n", queue_id);
2609 ds_put_cstr(ds, "Default:\n");
2612 SMAP_FOR_EACH (node, details) {
2613 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2616 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2618 if (stats.tx_packets != UINT64_MAX) {
2619 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2622 if (stats.tx_bytes != UINT64_MAX) {
2623 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2626 if (stats.tx_errors != UINT64_MAX) {
2627 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2630 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2631 queue_id, ovs_strerror(error));
2636 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2637 const char *argv[], void *aux OVS_UNUSED)
2639 struct ds ds = DS_EMPTY_INITIALIZER;
2640 struct smap smap = SMAP_INITIALIZER(&smap);
2641 struct iface *iface;
2643 struct smap_node *node;
2645 iface = iface_find(argv[1]);
2647 unixctl_command_reply_error(conn, "no such interface");
2651 netdev_get_qos(iface->netdev, &type, &smap);
2653 if (*type != '\0') {
2654 struct netdev_queue_dump dump;
2655 struct smap details;
2656 unsigned int queue_id;
2658 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2660 SMAP_FOR_EACH (node, &smap) {
2661 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2664 smap_init(&details);
2665 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2666 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2668 smap_destroy(&details);
2670 unixctl_command_reply(conn, ds_cstr(&ds));
2672 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2673 unixctl_command_reply_error(conn, ds_cstr(&ds));
2676 smap_destroy(&smap);
2680 /* Bridge reconfiguration functions. */
2682 bridge_create(const struct ovsrec_bridge *br_cfg)
2686 ovs_assert(!bridge_lookup(br_cfg->name));
2687 br = xzalloc(sizeof *br);
2689 br->name = xstrdup(br_cfg->name);
2690 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2693 /* Derive the default Ethernet address from the bridge's UUID. This should
2694 * be unique and it will be stable between ovs-vswitchd runs. */
2695 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2696 eth_addr_mark_random(br->default_ea);
2698 hmap_init(&br->ports);
2699 hmap_init(&br->ifaces);
2700 hmap_init(&br->iface_by_name);
2701 hmap_init(&br->mirrors);
2703 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2707 bridge_destroy(struct bridge *br)
2710 struct mirror *mirror, *next_mirror;
2711 struct port *port, *next_port;
2713 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2716 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2717 mirror_destroy(mirror);
2720 hmap_remove(&all_bridges, &br->node);
2721 ofproto_destroy(br->ofproto);
2722 hmap_destroy(&br->ifaces);
2723 hmap_destroy(&br->ports);
2724 hmap_destroy(&br->iface_by_name);
2725 hmap_destroy(&br->mirrors);
2732 static struct bridge *
2733 bridge_lookup(const char *name)
2737 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2738 if (!strcmp(br->name, name)) {
2745 /* Handle requests for a listing of all flows known by the OpenFlow
2746 * stack, including those normally hidden. */
2748 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2749 const char *argv[], void *aux OVS_UNUSED)
2754 br = bridge_lookup(argv[1]);
2756 unixctl_command_reply_error(conn, "Unknown bridge");
2761 ofproto_get_all_flows(br->ofproto, &results);
2763 unixctl_command_reply(conn, ds_cstr(&results));
2764 ds_destroy(&results);
2767 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2768 * connections and reconnect. If BRIDGE is not specified, then all bridges
2769 * drop their controller connections and reconnect. */
2771 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2772 const char *argv[], void *aux OVS_UNUSED)
2776 br = bridge_lookup(argv[1]);
2778 unixctl_command_reply_error(conn, "Unknown bridge");
2781 ofproto_reconnect_controllers(br->ofproto);
2783 HMAP_FOR_EACH (br, node, &all_bridges) {
2784 ofproto_reconnect_controllers(br->ofproto);
2787 unixctl_command_reply(conn, NULL);
2791 bridge_get_controllers(const struct bridge *br,
2792 struct ovsrec_controller ***controllersp)
2794 struct ovsrec_controller **controllers;
2795 size_t n_controllers;
2797 controllers = br->cfg->controller;
2798 n_controllers = br->cfg->n_controller;
2800 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2806 *controllersp = controllers;
2808 return n_controllers;
2812 bridge_collect_wanted_ports(struct bridge *br,
2813 const unsigned long int *splinter_vlans,
2814 struct shash *wanted_ports)
2818 shash_init(wanted_ports);
2820 for (i = 0; i < br->cfg->n_ports; i++) {
2821 const char *name = br->cfg->ports[i]->name;
2822 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2823 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2827 if (bridge_get_controllers(br, NULL)
2828 && !shash_find(wanted_ports, br->name)) {
2829 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2830 br->name, br->name);
2832 ovsrec_interface_init(&br->synth_local_iface);
2833 ovsrec_port_init(&br->synth_local_port);
2835 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2836 br->synth_local_port.n_interfaces = 1;
2837 br->synth_local_port.name = br->name;
2839 br->synth_local_iface.name = br->name;
2840 br->synth_local_iface.type = "internal";
2842 br->synth_local_ifacep = &br->synth_local_iface;
2844 shash_add(wanted_ports, br->name, &br->synth_local_port);
2847 if (splinter_vlans) {
2848 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2852 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2853 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2854 * 'br' needs to complete its configuration. */
2856 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2858 struct shash_node *port_node;
2859 struct port *port, *next;
2861 /* Get rid of deleted ports.
2862 * Get rid of deleted interfaces on ports that still exist. */
2863 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2864 port->cfg = shash_find_data(wanted_ports, port->name);
2868 port_del_ifaces(port);
2872 /* Update iface->cfg and iface->type in interfaces that still exist. */
2873 SHASH_FOR_EACH (port_node, wanted_ports) {
2874 const struct ovsrec_port *port = port_node->data;
2877 for (i = 0; i < port->n_interfaces; i++) {
2878 const struct ovsrec_interface *cfg = port->interfaces[i];
2879 struct iface *iface = iface_lookup(br, cfg->name);
2880 const char *type = iface_get_type(cfg, br->cfg);
2885 } else if (!strcmp(type, "null")) {
2886 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2887 " may be removed in February 2013. Please email"
2888 " dev@openvswitch.org with concerns.",
2891 /* We will add new interfaces later. */
2897 /* Initializes 'oc' appropriately as a management service controller for
2900 * The caller must free oc->target when it is no longer needed. */
2902 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2903 struct ofproto_controller *oc)
2905 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2906 oc->max_backoff = 0;
2907 oc->probe_interval = 60;
2908 oc->band = OFPROTO_OUT_OF_BAND;
2910 oc->burst_limit = 0;
2911 oc->enable_async_msgs = true;
2915 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2917 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2918 struct ofproto_controller *oc)
2922 oc->target = c->target;
2923 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2924 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2925 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2926 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2927 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2928 oc->burst_limit = (c->controller_burst_limit
2929 ? *c->controller_burst_limit : 0);
2930 oc->enable_async_msgs = (!c->enable_async_messages
2931 || *c->enable_async_messages);
2932 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2933 if (dscp < 0 || dscp > 63) {
2934 dscp = DSCP_DEFAULT;
2939 /* Configures the IP stack for 'br''s local interface properly according to the
2940 * configuration in 'c'. */
2942 bridge_configure_local_iface_netdev(struct bridge *br,
2943 struct ovsrec_controller *c)
2945 struct netdev *netdev;
2946 struct in_addr mask, gateway;
2948 struct iface *local_iface;
2951 /* If there's no local interface or no IP address, give up. */
2952 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2953 if (!local_iface || !c->local_ip
2954 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2958 /* Bring up the local interface. */
2959 netdev = local_iface->netdev;
2960 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2962 /* Configure the IP address and netmask. */
2963 if (!c->local_netmask
2964 || !inet_pton(AF_INET, c->local_netmask, &mask)
2966 mask.s_addr = guess_netmask(ip.s_addr);
2968 if (!netdev_set_in4(netdev, ip, mask)) {
2969 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2970 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2973 /* Configure the default gateway. */
2974 if (c->local_gateway
2975 && inet_pton(AF_INET, c->local_gateway, &gateway)
2976 && gateway.s_addr) {
2977 if (!netdev_add_router(netdev, gateway)) {
2978 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2979 br->name, IP_ARGS(gateway.s_addr));
2984 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2985 * in either string are treated as equal to any number of slashes in the other,
2986 * e.g. "x///y" is equal to "x/y".
2988 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2989 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2990 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2991 * 'b' against a prefix of 'a'.
2994 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2996 const char *b_start = b;
2998 if (b - b_start >= b_stoplen) {
3000 } else if (*a != *b) {
3002 } else if (*a == '/') {
3003 a += strspn(a, "/");
3004 b += strspn(b, "/");
3005 } else if (*a == '\0') {
3015 bridge_configure_remotes(struct bridge *br,
3016 const struct sockaddr_in *managers, size_t n_managers)
3018 bool disable_in_band;
3020 struct ovsrec_controller **controllers;
3021 size_t n_controllers;
3023 enum ofproto_fail_mode fail_mode;
3025 struct ofproto_controller *ocs;
3029 /* Check if we should disable in-band control on this bridge. */
3030 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3033 /* Set OpenFlow queue ID for in-band control. */
3034 ofproto_set_in_band_queue(br->ofproto,
3035 smap_get_int(&br->cfg->other_config,
3036 "in-band-queue", -1));
3038 if (disable_in_band) {
3039 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3041 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3044 n_controllers = bridge_get_controllers(br, &controllers);
3046 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3049 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3050 for (i = 0; i < n_controllers; i++) {
3051 struct ovsrec_controller *c = controllers[i];
3053 if (!strncmp(c->target, "punix:", 6)
3054 || !strncmp(c->target, "unix:", 5)) {
3055 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3058 if (!strncmp(c->target, "unix:", 5)) {
3059 /* Connect to a listening socket */
3060 whitelist = xasprintf("unix:%s/", ovs_rundir());
3061 if (strchr(c->target, '/') &&
3062 !equal_pathnames(c->target, whitelist,
3063 strlen(whitelist))) {
3064 /* Absolute path specified, but not in ovs_rundir */
3065 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3066 "controller \"%s\" due to possibility for "
3067 "remote exploit. Instead, specify socket "
3068 "in whitelisted \"%s\" or connect to "
3069 "\"unix:%s/%s.mgmt\" (which is always "
3070 "available without special configuration).",
3071 br->name, c->target, whitelist,
3072 ovs_rundir(), br->name);
3077 whitelist = xasprintf("punix:%s/%s.controller",
3078 ovs_rundir(), br->name);
3079 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3080 /* Prevent remote ovsdb-server users from accessing
3081 * arbitrary Unix domain sockets and overwriting arbitrary
3083 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3084 "controller \"%s\" due to possibility of "
3085 "overwriting local files. Instead, specify "
3086 "whitelisted \"%s\" or connect to "
3087 "\"unix:%s/%s.mgmt\" (which is always "
3088 "available without special configuration).",
3089 br->name, c->target, whitelist,
3090 ovs_rundir(), br->name);
3099 bridge_configure_local_iface_netdev(br, c);
3100 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3101 if (disable_in_band) {
3102 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3107 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3108 bridge_get_allowed_versions(br));
3109 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3112 /* Set the fail-mode. */
3113 fail_mode = !br->cfg->fail_mode
3114 || !strcmp(br->cfg->fail_mode, "standalone")
3115 ? OFPROTO_FAIL_STANDALONE
3116 : OFPROTO_FAIL_SECURE;
3117 ofproto_set_fail_mode(br->ofproto, fail_mode);
3119 /* Configure OpenFlow controller connection snooping. */
3120 if (!ofproto_has_snoops(br->ofproto)) {
3124 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3125 ovs_rundir(), br->name));
3126 ofproto_set_snoops(br->ofproto, &snoops);
3127 sset_destroy(&snoops);
3132 bridge_configure_tables(struct bridge *br)
3134 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3138 n_tables = ofproto_get_n_tables(br->ofproto);
3140 for (i = 0; i < n_tables; i++) {
3141 struct ofproto_table_settings s;
3142 bool use_default_prefixes = true;
3145 s.max_flows = UINT_MAX;
3148 s.n_prefix_fields = 0;
3149 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3151 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3152 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3155 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3156 s.max_flows = *cfg->flow_limit;
3158 if (cfg->overflow_policy
3159 && !strcmp(cfg->overflow_policy, "evict")) {
3161 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3162 for (k = 0; k < cfg->n_groups; k++) {
3163 const char *string = cfg->groups[k];
3166 msg = mf_parse_subfield__(&s.groups[k], &string);
3168 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3169 "'groups' (%s)", br->name, i, msg);
3171 } else if (*string) {
3172 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3173 "element '%s' contains trailing garbage",
3174 br->name, i, cfg->groups[k]);
3180 /* Prefix lookup fields. */
3181 s.n_prefix_fields = 0;
3182 for (k = 0; k < cfg->n_prefixes; k++) {
3183 const char *name = cfg->prefixes[k];
3184 const struct mf_field *mf;
3186 if (strcmp(name, "none") == 0) {
3187 use_default_prefixes = false;
3188 s.n_prefix_fields = 0;
3191 mf = mf_from_name(name);
3193 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3197 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3198 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3199 "%s", br->name, name);
3202 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3203 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3204 "field not used: %s", br->name, name);
3207 use_default_prefixes = false;
3208 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3211 if (use_default_prefixes) {
3212 /* Use default values. */
3213 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3214 memcpy(s.prefix_fields, default_prefix_fields,
3215 sizeof default_prefix_fields);
3218 struct ds ds = DS_EMPTY_INITIALIZER;
3219 for (k = 0; k < s.n_prefix_fields; k++) {
3221 ds_put_char(&ds, ',');
3223 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3225 if (s.n_prefix_fields == 0) {
3226 ds_put_cstr(&ds, "none");
3228 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3229 br->name, i, ds_cstr(&ds));
3233 ofproto_configure_table(br->ofproto, i, &s);
3237 for (; j < br->cfg->n_flow_tables; j++) {
3238 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3239 "%"PRId64" not supported by this datapath", br->name,
3240 br->cfg->key_flow_tables[j]);
3245 bridge_configure_dp_desc(struct bridge *br)
3247 ofproto_set_dp_desc(br->ofproto,
3248 smap_get(&br->cfg->other_config, "dp-desc"));
3251 /* Port functions. */
3253 static struct port *
3254 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3258 port = xzalloc(sizeof *port);
3260 port->name = xstrdup(cfg->name);
3262 list_init(&port->ifaces);
3264 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3268 /* Deletes interfaces from 'port' that are no longer configured for it. */
3270 port_del_ifaces(struct port *port)
3272 struct iface *iface, *next;
3273 struct sset new_ifaces;
3276 /* Collect list of new interfaces. */
3277 sset_init(&new_ifaces);
3278 for (i = 0; i < port->cfg->n_interfaces; i++) {
3279 const char *name = port->cfg->interfaces[i]->name;
3280 const char *type = port->cfg->interfaces[i]->type;
3281 if (strcmp(type, "null")) {
3282 sset_add(&new_ifaces, name);
3286 /* Get rid of deleted interfaces. */
3287 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3288 if (!sset_contains(&new_ifaces, iface->name)) {
3289 iface_destroy(iface);
3293 sset_destroy(&new_ifaces);
3297 port_destroy(struct port *port)
3300 struct bridge *br = port->bridge;
3301 struct iface *iface, *next;
3304 ofproto_bundle_unregister(br->ofproto, port);
3307 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3308 iface_destroy__(iface);
3311 hmap_remove(&br->ports, &port->hmap_node);
3317 static struct port *
3318 port_lookup(const struct bridge *br, const char *name)
3322 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3324 if (!strcmp(port->name, name)) {
3332 enable_lacp(struct port *port, bool *activep)
3334 if (!port->cfg->lacp) {
3335 /* XXX when LACP implementation has been sufficiently tested, enable by
3336 * default and make active on bonded ports. */
3338 } else if (!strcmp(port->cfg->lacp, "off")) {
3340 } else if (!strcmp(port->cfg->lacp, "active")) {
3343 } else if (!strcmp(port->cfg->lacp, "passive")) {
3347 VLOG_WARN("port %s: unknown LACP mode %s",
3348 port->name, port->cfg->lacp);
3353 static struct lacp_settings *
3354 port_configure_lacp(struct port *port, struct lacp_settings *s)
3356 const char *lacp_time, *system_id;
3359 if (!enable_lacp(port, &s->active)) {
3363 s->name = port->name;
3365 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3367 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3368 ETH_ADDR_SCAN_ARGS(s->id))) {
3369 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3370 " address.", port->name, system_id);
3374 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3377 if (eth_addr_is_zero(s->id)) {
3378 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3382 /* Prefer bondable links if unspecified. */
3383 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3385 s->priority = (priority > 0 && priority <= UINT16_MAX
3387 : UINT16_MAX - !list_is_short(&port->ifaces));
3389 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3390 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3392 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3393 "lacp-fallback-ab", false);
3399 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3401 int priority, portid, key;
3403 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3404 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3406 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3408 if (portid <= 0 || portid > UINT16_MAX) {
3409 portid = ofp_to_u16(iface->ofp_port);
3412 if (priority <= 0 || priority > UINT16_MAX) {
3413 priority = UINT16_MAX;
3416 if (key < 0 || key > UINT16_MAX) {
3420 s->name = iface->name;
3422 s->priority = priority;
3427 port_configure_bond(struct port *port, struct bond_settings *s)
3429 const char *detect_s;
3430 struct iface *iface;
3431 int miimon_interval;
3433 s->name = port->name;
3435 if (port->cfg->bond_mode) {
3436 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3437 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3438 port->name, port->cfg->bond_mode,
3439 bond_mode_to_string(s->balance));
3442 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3444 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3445 * active-backup. At some point we should remove this warning. */
3446 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3447 " in previous versions, the default bond_mode was"
3448 " balance-slb", port->name,
3449 bond_mode_to_string(s->balance));
3451 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3452 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3453 "please use another bond type or disable flood_vlans",
3457 miimon_interval = smap_get_int(&port->cfg->other_config,
3458 "bond-miimon-interval", 0);
3459 if (miimon_interval <= 0) {
3460 miimon_interval = 200;
3463 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3464 if (!detect_s || !strcmp(detect_s, "carrier")) {
3465 miimon_interval = 0;
3466 } else if (strcmp(detect_s, "miimon")) {
3467 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3468 "defaulting to carrier", port->name, detect_s);
3469 miimon_interval = 0;
3472 s->up_delay = MAX(0, port->cfg->bond_updelay);
3473 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3474 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3475 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3476 "bond-rebalance-interval", 10000);
3477 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3478 s->rebalance_interval = 1000;
3481 s->fake_iface = port->cfg->bond_fake_iface;
3483 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3484 "lacp-fallback-ab", false);
3486 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3487 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3491 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3492 * instead of obtaining it from the database. */
3494 port_is_synthetic(const struct port *port)
3496 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3499 /* Interface functions. */
3502 iface_is_internal(const struct ovsrec_interface *iface,
3503 const struct ovsrec_bridge *br)
3505 /* The local port and "internal" ports are always "internal". */
3506 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3509 /* Returns the correct network device type for interface 'iface' in bridge
3512 iface_get_type(const struct ovsrec_interface *iface,
3513 const struct ovsrec_bridge *br)
3517 /* The local port always has type "internal". Other ports take
3518 * their type from the database and default to "system" if none is
3520 if (iface_is_internal(iface, br)) {
3523 type = iface->type[0] ? iface->type : "system";
3526 return ofproto_port_open_type(br->datapath_type, type);
3530 iface_destroy__(struct iface *iface)
3533 struct port *port = iface->port;
3534 struct bridge *br = port->bridge;
3536 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3537 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3540 if (iface->ofp_port != OFPP_NONE) {
3541 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3544 list_remove(&iface->port_elem);
3545 hmap_remove(&br->iface_by_name, &iface->name_node);
3547 /* The user is changing configuration here, so netdev_remove needs to be
3548 * used as opposed to netdev_close */
3549 netdev_remove(iface->netdev);
3557 iface_destroy(struct iface *iface)
3560 struct port *port = iface->port;
3562 iface_destroy__(iface);
3563 if (list_is_empty(&port->ifaces)) {
3569 static struct iface *
3570 iface_lookup(const struct bridge *br, const char *name)
3572 struct iface *iface;
3574 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3575 &br->iface_by_name) {
3576 if (!strcmp(iface->name, name)) {
3584 static struct iface *
3585 iface_find(const char *name)
3587 const struct bridge *br;
3589 HMAP_FOR_EACH (br, node, &all_bridges) {
3590 struct iface *iface = iface_lookup(br, name);
3599 static struct iface *
3600 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3602 struct iface *iface;
3604 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3606 if (iface->ofp_port == ofp_port) {
3613 /* Set Ethernet address of 'iface', if one is specified in the configuration
3616 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3618 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3619 struct iface *hw_addr_iface;
3621 if (strcmp(iface->type, "internal")) {
3625 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3627 } else if (port->cfg->fake_bridge) {
3628 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3629 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3634 if (iface->ofp_port == OFPP_LOCAL) {
3635 VLOG_ERR("interface %s: ignoring mac in Interface record "
3636 "(use Bridge record to set local port's mac)",
3638 } else if (eth_addr_is_multicast(mac)) {
3639 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3642 int error = netdev_set_etheraddr(iface->netdev, mac);
3644 VLOG_ERR("interface %s: setting MAC failed (%s)",
3645 iface->name, ovs_strerror(error));
3651 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3653 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3655 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3656 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3657 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3661 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3662 * sets the "ofport" field to -1.
3664 * This is appropriate when 'if_cfg''s interface cannot be created or is
3665 * otherwise invalid. */
3667 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
3669 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3670 iface_set_ofport(if_cfg, OFPP_NONE);
3671 ovsrec_interface_set_error(if_cfg, errp);
3672 ovsrec_interface_set_status(if_cfg, NULL);
3673 ovsrec_interface_set_admin_state(if_cfg, NULL);
3674 ovsrec_interface_set_duplex(if_cfg, NULL);
3675 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3676 ovsrec_interface_set_link_state(if_cfg, NULL);
3677 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3678 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3679 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3680 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3681 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3682 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3683 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3684 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3689 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3691 union ovsdb_atom atom;
3693 atom.integer = target;
3694 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3698 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3700 struct ofpbuf queues_buf;
3702 ofpbuf_init(&queues_buf, 0);
3704 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3705 netdev_set_qos(iface->netdev, NULL, NULL);
3707 const struct ovsdb_datum *queues;
3708 struct netdev_queue_dump dump;
3709 unsigned int queue_id;
3710 struct smap details;
3714 /* Configure top-level Qos for 'iface'. */
3715 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3717 /* Deconfigure queues that were deleted. */
3718 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3720 smap_init(&details);
3721 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3722 if (!queue_ids_include(queues, queue_id)) {
3723 netdev_delete_queue(iface->netdev, queue_id);
3726 smap_destroy(&details);
3728 /* Configure queues for 'iface'. */
3730 for (i = 0; i < qos->n_queues; i++) {
3731 const struct ovsrec_queue *queue = qos->value_queues[i];
3732 unsigned int queue_id = qos->key_queues[i];
3734 if (queue_id == 0) {
3738 if (queue->n_dscp == 1) {
3739 struct ofproto_port_queue *port_queue;
3741 port_queue = ofpbuf_put_uninit(&queues_buf,
3742 sizeof *port_queue);
3743 port_queue->queue = queue_id;
3744 port_queue->dscp = queue->dscp[0];
3747 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3750 struct smap details;
3752 smap_init(&details);
3753 netdev_set_queue(iface->netdev, 0, &details);
3754 smap_destroy(&details);
3758 if (iface->ofp_port != OFPP_NONE) {
3759 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3760 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3762 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3763 port_queues, n_queues);
3766 netdev_set_policing(iface->netdev,
3767 iface->cfg->ingress_policing_rate,
3768 iface->cfg->ingress_policing_burst);
3770 ofpbuf_uninit(&queues_buf);
3774 iface_configure_cfm(struct iface *iface)
3776 const struct ovsrec_interface *cfg = iface->cfg;
3777 const char *opstate_str;
3778 const char *cfm_ccm_vlan;
3779 struct cfm_settings s;
3780 struct smap netdev_args;
3782 if (!cfg->n_cfm_mpid) {
3783 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3787 s.check_tnl_key = false;
3788 smap_init(&netdev_args);
3789 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3790 const char *key = smap_get(&netdev_args, "key");
3791 const char *in_key = smap_get(&netdev_args, "in_key");
3793 s.check_tnl_key = (key && !strcmp(key, "flow"))
3794 || (in_key && !strcmp(in_key, "flow"));
3796 smap_destroy(&netdev_args);
3798 s.mpid = *cfg->cfm_mpid;
3799 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3800 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3801 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3803 if (s.interval <= 0) {
3807 if (!cfm_ccm_vlan) {
3809 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3810 s.ccm_vlan = CFM_RANDOM_VLAN;
3812 s.ccm_vlan = atoi(cfm_ccm_vlan);
3813 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3818 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3820 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3822 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3823 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3825 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3828 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3829 * instead of obtaining it from the database. */
3831 iface_is_synthetic(const struct iface *iface)
3833 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3837 iface_validate_ofport__(size_t n, int64_t *ofport)
3839 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3840 ? u16_to_ofp(*ofport)
3845 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3847 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3851 iface_pick_ofport(const struct ovsrec_interface *cfg)
3853 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3854 return (requested_ofport != OFPP_NONE
3856 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3859 /* Port mirroring. */
3861 static struct mirror *
3862 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3866 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3867 if (uuid_equals(uuid, &m->uuid)) {
3875 bridge_configure_mirrors(struct bridge *br)
3877 const struct ovsdb_datum *mc;
3878 unsigned long *flood_vlans;
3879 struct mirror *m, *next;
3882 /* Get rid of deleted mirrors. */
3883 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3884 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3885 union ovsdb_atom atom;
3887 atom.uuid = m->uuid;
3888 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3893 /* Add new mirrors and reconfigure existing ones. */
3894 for (i = 0; i < br->cfg->n_mirrors; i++) {
3895 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3896 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3898 m = mirror_create(br, cfg);
3901 if (!mirror_configure(m)) {
3906 /* Update flooded vlans (for RSPAN). */
3907 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3908 br->cfg->n_flood_vlans);
3909 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3910 bitmap_free(flood_vlans);
3913 static struct mirror *
3914 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3918 m = xzalloc(sizeof *m);
3919 m->uuid = cfg->header_.uuid;
3920 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3922 m->name = xstrdup(cfg->name);
3928 mirror_destroy(struct mirror *m)
3931 struct bridge *br = m->bridge;
3934 ofproto_mirror_unregister(br->ofproto, m);
3937 hmap_remove(&br->mirrors, &m->hmap_node);
3944 mirror_collect_ports(struct mirror *m,
3945 struct ovsrec_port **in_ports, int n_in_ports,
3946 void ***out_portsp, size_t *n_out_portsp)
3948 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3949 size_t n_out_ports = 0;
3952 for (i = 0; i < n_in_ports; i++) {
3953 const char *name = in_ports[i]->name;
3954 struct port *port = port_lookup(m->bridge, name);
3956 out_ports[n_out_ports++] = port;
3958 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3959 "port %s", m->bridge->name, m->name, name);
3962 *out_portsp = out_ports;
3963 *n_out_portsp = n_out_ports;
3967 mirror_configure(struct mirror *m)
3969 const struct ovsrec_mirror *cfg = m->cfg;
3970 struct ofproto_mirror_settings s;
3973 if (strcmp(cfg->name, m->name)) {
3975 m->name = xstrdup(cfg->name);
3979 /* Get output port or VLAN. */
3980 if (cfg->output_port) {
3981 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3982 if (!s.out_bundle) {
3983 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3984 m->bridge->name, m->name);
3987 s.out_vlan = UINT16_MAX;
3989 if (cfg->output_vlan) {
3990 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3991 "output vlan; ignoring output vlan",
3992 m->bridge->name, m->name);
3994 } else if (cfg->output_vlan) {
3995 /* The database should prevent invalid VLAN values. */
3996 s.out_bundle = NULL;
3997 s.out_vlan = *cfg->output_vlan;
3999 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4000 m->bridge->name, m->name);
4004 /* Get port selection. */
4005 if (cfg->select_all) {
4006 size_t n_ports = hmap_count(&m->bridge->ports);
4007 void **ports = xmalloc(n_ports * sizeof *ports);
4012 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4022 /* Get ports, dropping ports that don't exist.
4023 * The IDL ensures that there are no duplicates. */
4024 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4025 &s.srcs, &s.n_srcs);
4026 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4027 &s.dsts, &s.n_dsts);
4030 /* Get VLAN selection. */
4031 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4034 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4037 if (s.srcs != s.dsts) {
4046 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4048 * This is deprecated. It is only for compatibility with broken device drivers
4049 * in old versions of Linux that do not properly support VLANs when VLAN
4050 * devices are not used. When broken device drivers are no longer in
4051 * widespread use, we will delete these interfaces. */
4053 static struct ovsrec_port **recs;
4054 static size_t n_recs, allocated_recs;
4056 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4057 * splinters are reconfigured. */
4059 register_rec(struct ovsrec_port *rec)
4061 if (n_recs >= allocated_recs) {
4062 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4064 recs[n_recs++] = rec;
4067 /* Frees all of the ports registered with register_reg(). */
4069 free_registered_recs(void)
4073 for (i = 0; i < n_recs; i++) {
4074 struct ovsrec_port *port = recs[i];
4077 for (j = 0; j < port->n_interfaces; j++) {
4078 struct ovsrec_interface *iface = port->interfaces[j];
4083 smap_destroy(&port->other_config);
4084 free(port->interfaces);
4092 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4095 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4097 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4101 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4104 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4105 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4106 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4109 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4110 * use, returns NULL.
4112 * Updates 'vlan_splinters_enabled_anywhere'. */
4113 static unsigned long int *
4114 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4116 unsigned long int *splinter_vlans;
4117 struct sset splinter_ifaces;
4118 const char *real_dev_name;
4119 struct shash *real_devs;
4120 struct shash_node *node;
4124 /* Free space allocated for synthesized ports and interfaces, since we're
4125 * in the process of reconstructing all of them. */
4126 free_registered_recs();
4128 splinter_vlans = bitmap_allocate(4096);
4129 sset_init(&splinter_ifaces);
4130 vlan_splinters_enabled_anywhere = false;
4131 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4132 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4135 for (j = 0; j < br_cfg->n_ports; j++) {
4136 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4139 for (k = 0; k < port_cfg->n_interfaces; k++) {
4140 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4142 if (vlan_splinters_is_enabled(iface_cfg)) {
4143 vlan_splinters_enabled_anywhere = true;
4144 sset_add(&splinter_ifaces, iface_cfg->name);
4145 vlan_bitmap_from_array__(port_cfg->trunks,
4151 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4152 bitmap_set1(splinter_vlans, *port_cfg->tag);
4157 if (!vlan_splinters_enabled_anywhere) {
4158 free(splinter_vlans);
4159 sset_destroy(&splinter_ifaces);
4163 HMAP_FOR_EACH (br, node, &all_bridges) {
4165 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4169 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4170 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4171 * device to be created for it. */
4172 bitmap_set0(splinter_vlans, 0);
4173 bitmap_set0(splinter_vlans, 4095);
4175 /* Delete all VLAN devices that we don't need. */
4177 real_devs = vlandev_get_real_devs();
4178 SHASH_FOR_EACH (node, real_devs) {
4179 const struct vlan_real_dev *real_dev = node->data;
4180 const struct vlan_dev *vlan_dev;
4181 bool real_dev_has_splinters;
4183 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4185 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4186 if (!real_dev_has_splinters
4187 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4188 struct netdev *netdev;
4190 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4191 if (!netdev_get_in4(netdev, NULL, NULL) ||
4192 !netdev_get_in6(netdev, NULL)) {
4193 /* It has an IP address configured, so we don't own
4194 * it. Don't delete it. */
4196 vlandev_del(vlan_dev->name);
4198 netdev_close(netdev);
4205 /* Add all VLAN devices that we need. */
4206 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4209 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4210 if (!vlandev_get_name(real_dev_name, vid)) {
4211 vlandev_add(real_dev_name, vid);
4218 sset_destroy(&splinter_ifaces);
4220 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4221 free(splinter_vlans);
4224 return splinter_vlans;
4227 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4230 configure_splinter_port(struct port *port)
4232 struct ofproto *ofproto = port->bridge->ofproto;
4233 ofp_port_t realdev_ofp_port;
4234 const char *realdev_name;
4235 struct iface *vlandev, *realdev;
4237 ofproto_bundle_unregister(port->bridge->ofproto, port);
4239 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4242 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4243 realdev = iface_lookup(port->bridge, realdev_name);
4244 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4246 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4250 static struct ovsrec_port *
4251 synthesize_splinter_port(const char *real_dev_name,
4252 const char *vlan_dev_name, int vid)
4254 struct ovsrec_interface *iface;
4255 struct ovsrec_port *port;
4257 iface = xmalloc(sizeof *iface);
4258 ovsrec_interface_init(iface);
4259 iface->name = xstrdup(vlan_dev_name);
4260 iface->type = "system";
4262 port = xmalloc(sizeof *port);
4263 ovsrec_port_init(port);
4264 port->interfaces = xmemdup(&iface, sizeof iface);
4265 port->n_interfaces = 1;
4266 port->name = xstrdup(vlan_dev_name);
4267 port->vlan_mode = "splinter";
4268 port->tag = xmalloc(sizeof *port->tag);
4271 smap_add(&port->other_config, "realdev", real_dev_name);
4277 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4278 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4279 * 1-bit in the 'splinter_vlans' bitmap. */
4281 add_vlan_splinter_ports(struct bridge *br,
4282 const unsigned long int *splinter_vlans,
4283 struct shash *ports)
4287 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4288 * we're modifying 'ports'. */
4289 for (i = 0; i < br->cfg->n_ports; i++) {
4290 const char *name = br->cfg->ports[i]->name;
4291 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4294 for (j = 0; j < port_cfg->n_interfaces; j++) {
4295 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4297 if (vlan_splinters_is_enabled(iface_cfg)) {
4298 const char *real_dev_name;
4301 real_dev_name = iface_cfg->name;
4302 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4303 const char *vlan_dev_name;
4305 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4307 && !shash_find(ports, vlan_dev_name)) {
4308 shash_add(ports, vlan_dev_name,
4309 synthesize_splinter_port(
4310 real_dev_name, vlan_dev_name, vid));
4319 mirror_refresh_stats(struct mirror *m)
4321 struct ofproto *ofproto = m->bridge->ofproto;
4322 uint64_t tx_packets, tx_bytes;
4325 size_t stat_cnt = 0;
4327 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4328 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4332 if (tx_packets != UINT64_MAX) {
4333 keys[stat_cnt] = "tx_packets";
4334 values[stat_cnt] = tx_packets;
4337 if (tx_bytes != UINT64_MAX) {
4338 keys[stat_cnt] = "tx_bytes";
4339 values[stat_cnt] = tx_bytes;
4343 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);