1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
65 VLOG_DEFINE_THIS_MODULE(bridge);
67 COVERAGE_DEFINE(bridge_reconfigure);
70 /* These members are always valid.
72 * They are immutable: they never change between iface_create() and
74 struct list port_elem; /* Element in struct port's "ifaces" list. */
75 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
76 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
77 struct port *port; /* Containing port. */
78 char *name; /* Host network device name. */
79 struct netdev *netdev; /* Network device. */
80 ofp_port_t ofp_port; /* OpenFlow port number. */
83 /* These members are valid only within bridge_reconfigure(). */
84 const char *type; /* Usually same as cfg->type. */
85 const struct ovsrec_interface *cfg;
89 struct uuid uuid; /* UUID of this "mirror" record in database. */
90 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
91 struct bridge *bridge;
93 const struct ovsrec_mirror *cfg;
97 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
98 struct bridge *bridge;
101 const struct ovsrec_port *cfg;
103 /* An ordinary bridge port has 1 interface.
104 * A bridge port for bonding has at least 2 interfaces. */
105 struct list ifaces; /* List of "struct iface"s. */
109 struct hmap_node node; /* In 'all_bridges'. */
110 char *name; /* User-specified arbitrary name. */
111 char *type; /* Datapath type. */
112 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
113 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
114 const struct ovsrec_bridge *cfg;
116 /* OpenFlow switch processing. */
117 struct ofproto *ofproto; /* OpenFlow switch. */
120 struct hmap ports; /* "struct port"s indexed by name. */
121 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
122 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
124 /* Port mirroring. */
125 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
127 /* Used during reconfiguration. */
128 struct shash wanted_ports;
130 /* Synthetic local port if necessary. */
131 struct ovsrec_port synth_local_port;
132 struct ovsrec_interface synth_local_iface;
133 struct ovsrec_interface *synth_local_ifacep;
136 /* All bridges, indexed by name. */
137 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
139 /* OVSDB IDL used to obtain configuration. */
140 static struct ovsdb_idl *idl;
142 /* We want to complete daemonization, fully detaching from our parent process,
143 * only after we have completed our initial configuration, committed our state
144 * to the database, and received confirmation back from the database server
145 * that it applied the commit. This allows our parent process to know that,
146 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
147 * to have already been filled in. (It is only "very likely" rather than
148 * certain because there is always a slim possibility that the transaction will
149 * fail or that some other client has added new bridges, ports, etc. while
150 * ovs-vswitchd was configuring using an old configuration.)
152 * We only need to do this once for our initial configuration at startup, so
153 * 'initial_config_done' tracks whether we've already done it. While we are
154 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
155 * itself and is otherwise NULL. */
156 static bool initial_config_done;
157 static struct ovsdb_idl_txn *daemonize_txn;
159 /* Most recently processed IDL sequence number. */
160 static unsigned int idl_seqno;
162 /* Track changes to port connectivity. */
163 static uint64_t connectivity_seqno = LLONG_MIN;
165 /* Status update to database.
167 * Some information in the database must be kept as up-to-date as possible to
168 * allow controllers to respond rapidly to network outages. Those status are
169 * updated via the 'status_txn'.
171 * We use the global connectivity sequence number to detect the status change.
172 * Also, to prevent the status update from sending too much to the database,
173 * we check the return status of each update transaction and do not start new
174 * update if the previous transaction status is 'TXN_INCOMPLETE'.
176 * 'status_txn' is NULL if there is no ongoing status update.
178 * If the previous database transaction was incomplete or failed (is not
179 * 'TXN_SUCCESS' or 'TXN_UNCHANGED'), 'force_status_commit' is set to true.
180 * This means that 'status_txn' must be committed next iteration of bridge_run()
181 * even if the connectivity or netdev sequence numbers do not change.
183 static struct ovsdb_idl_txn *status_txn;
184 static bool force_status_commit = true;
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 /* Current stats database transaction, NULL if there is no ongoing
197 static struct ovsdb_idl_txn *stats_txn;
199 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
200 * expensive. This can cause bridge_reconfigure() to take a long time during
201 * which no other work can be done. To deal with this problem, we limit port
202 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
203 * call to bridge_reconfigure(). If there is more work to do after the limit
204 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
205 * This allows the rest of the code to catch up on important things like
206 * forwarding packets. */
207 #define OFP_PORT_ACTION_WINDOW 10
209 static void add_del_bridges(const struct ovsrec_open_vswitch *);
210 static void bridge_run__(void);
211 static void bridge_create(const struct ovsrec_bridge *);
212 static void bridge_destroy(struct bridge *);
213 static struct bridge *bridge_lookup(const char *name);
214 static unixctl_cb_func bridge_unixctl_dump_flows;
215 static unixctl_cb_func bridge_unixctl_reconnect;
216 static size_t bridge_get_controllers(const struct bridge *br,
217 struct ovsrec_controller ***controllersp);
218 static void bridge_collect_wanted_ports(struct bridge *,
219 const unsigned long *splinter_vlans,
220 struct shash *wanted_ports);
221 static void bridge_delete_ofprotos(void);
222 static void bridge_delete_or_reconfigure_ports(struct bridge *);
223 static void bridge_del_ports(struct bridge *,
224 const struct shash *wanted_ports);
225 static void bridge_add_ports(struct bridge *,
226 const struct shash *wanted_ports);
228 static void bridge_configure_datapath_id(struct bridge *);
229 static void bridge_configure_netflow(struct bridge *);
230 static void bridge_configure_forward_bpdu(struct bridge *);
231 static void bridge_configure_mac_table(struct bridge *);
232 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
233 static void bridge_configure_ipfix(struct bridge *);
234 static void bridge_configure_stp(struct bridge *);
235 static void bridge_configure_tables(struct bridge *);
236 static void bridge_configure_dp_desc(struct bridge *);
237 static void bridge_configure_remotes(struct bridge *,
238 const struct sockaddr_in *managers,
240 static void bridge_pick_local_hw_addr(struct bridge *,
241 uint8_t ea[ETH_ADDR_LEN],
242 struct iface **hw_addr_iface);
243 static uint64_t bridge_pick_datapath_id(struct bridge *,
244 const uint8_t bridge_ea[ETH_ADDR_LEN],
245 struct iface *hw_addr_iface);
246 static uint64_t dpid_from_hash(const void *, size_t nbytes);
247 static bool bridge_has_bond_fake_iface(const struct bridge *,
249 static bool port_is_bond_fake_iface(const struct port *);
251 static unixctl_cb_func qos_unixctl_show;
253 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
254 static void port_del_ifaces(struct port *);
255 static void port_destroy(struct port *);
256 static struct port *port_lookup(const struct bridge *, const char *name);
257 static void port_configure(struct port *);
258 static struct lacp_settings *port_configure_lacp(struct port *,
259 struct lacp_settings *);
260 static void port_configure_bond(struct port *, struct bond_settings *);
261 static bool port_is_synthetic(const struct port *);
263 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
264 static void run_system_stats(void);
266 static void bridge_configure_mirrors(struct bridge *);
267 static struct mirror *mirror_create(struct bridge *,
268 const struct ovsrec_mirror *);
269 static void mirror_destroy(struct mirror *);
270 static bool mirror_configure(struct mirror *);
271 static void mirror_refresh_stats(struct mirror *);
273 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
274 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
275 const struct ovsrec_port *);
276 static bool iface_is_internal(const struct ovsrec_interface *iface,
277 const struct ovsrec_bridge *br);
278 static const char *iface_get_type(const struct ovsrec_interface *,
279 const struct ovsrec_bridge *);
280 static void iface_destroy(struct iface *);
281 static void iface_destroy__(struct iface *);
282 static struct iface *iface_lookup(const struct bridge *, const char *name);
283 static struct iface *iface_find(const char *name);
284 static struct iface *iface_from_ofp_port(const struct bridge *,
285 ofp_port_t ofp_port);
286 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
287 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
288 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
289 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
290 static void iface_configure_cfm(struct iface *);
291 static void iface_refresh_cfm_stats(struct iface *);
292 static void iface_refresh_stats(struct iface *);
293 static void iface_refresh_netdev_status(struct iface *);
294 static void iface_refresh_ofproto_status(struct iface *);
295 static bool iface_is_synthetic(const struct iface *);
296 static ofp_port_t iface_get_requested_ofp_port(
297 const struct ovsrec_interface *);
298 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
300 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
302 * This is deprecated. It is only for compatibility with broken device drivers
303 * in old versions of Linux that do not properly support VLANs when VLAN
304 * devices are not used. When broken device drivers are no longer in
305 * widespread use, we will delete these interfaces. */
307 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
308 static bool vlan_splinters_enabled_anywhere;
310 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
311 static unsigned long int *collect_splinter_vlans(
312 const struct ovsrec_open_vswitch *);
313 static void configure_splinter_port(struct port *);
314 static void add_vlan_splinter_ports(struct bridge *,
315 const unsigned long int *splinter_vlans,
316 struct shash *ports);
319 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
321 struct shash iface_hints;
322 static bool initialized = false;
329 shash_init(&iface_hints);
332 for (i = 0; i < cfg->n_bridges; i++) {
333 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
336 for (j = 0; j < br_cfg->n_ports; j++) {
337 struct ovsrec_port *port_cfg = br_cfg->ports[j];
340 for (k = 0; k < port_cfg->n_interfaces; k++) {
341 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
342 struct iface_hint *iface_hint;
344 iface_hint = xmalloc(sizeof *iface_hint);
345 iface_hint->br_name = br_cfg->name;
346 iface_hint->br_type = br_cfg->datapath_type;
347 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
349 shash_add(&iface_hints, if_cfg->name, iface_hint);
355 ofproto_init(&iface_hints);
357 shash_destroy_free_data(&iface_hints);
361 /* Public functions. */
363 /* Initializes the bridge module, configuring it to obtain its configuration
364 * from an OVSDB server accessed over 'remote', which should be a string in a
365 * form acceptable to ovsdb_idl_create(). */
367 bridge_init(const char *remote)
369 /* Create connection to database. */
370 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
371 idl_seqno = ovsdb_idl_get_seqno(idl);
372 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
373 ovsdb_idl_verify_write_only(idl);
375 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
376 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
378 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
379 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
380 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
381 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
383 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
384 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
385 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
387 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
388 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
389 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
390 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
408 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
409 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
410 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
411 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
413 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
414 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
415 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
416 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
418 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
420 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
422 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
423 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
425 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
426 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
427 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
428 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
430 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
431 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
432 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
433 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
434 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
436 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
438 /* Register unixctl commands. */
439 unixctl_command_register("qos/show", "interface", 1, 1,
440 qos_unixctl_show, NULL);
441 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
442 bridge_unixctl_dump_flows, NULL);
443 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
444 bridge_unixctl_reconnect, NULL);
454 struct bridge *br, *next_br;
456 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
459 ovsdb_idl_destroy(idl);
462 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
463 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
464 * responsible for freeing '*managersp' (with free()).
466 * You may be asking yourself "why does ovs-vswitchd care?", because
467 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
468 * should not be and in fact is not directly involved in that. But
469 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
470 * it has to tell in-band control where the managers are to enable that.
471 * (Thus, only managers connected in-band are collected.)
474 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
475 struct sockaddr_in **managersp, size_t *n_managersp)
477 struct sockaddr_in *managers = NULL;
478 size_t n_managers = 0;
482 /* Collect all of the potential targets from the "targets" columns of the
483 * rows pointed to by "manager_options", excluding any that are
486 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
487 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
489 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
490 sset_find_and_delete(&targets, m->target);
492 sset_add(&targets, m->target);
496 /* Now extract the targets' IP addresses. */
497 if (!sset_is_empty(&targets)) {
500 managers = xmalloc(sset_count(&targets) * sizeof *managers);
501 SSET_FOR_EACH (target, &targets) {
502 struct sockaddr_storage ss;
504 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
506 && ss.ss_family == AF_INET) {
507 managers[n_managers++] = *(struct sockaddr_in *) &ss;
511 sset_destroy(&targets);
513 *managersp = managers;
514 *n_managersp = n_managers;
518 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
520 unsigned long int *splinter_vlans;
521 struct sockaddr_in *managers;
522 struct bridge *br, *next;
523 int sflow_bridge_number;
526 COVERAGE_INC(bridge_reconfigure);
528 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
529 OFPROTO_FLOW_LIMIT_DEFAULT));
530 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
531 OFPROTO_MAX_IDLE_DEFAULT));
534 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
535 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
537 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
538 * to 'ovs_cfg', with only very minimal configuration otherwise.
540 * This is mostly an update to bridge data structures. Nothing is pushed
541 * down to ofproto or lower layers. */
542 add_del_bridges(ovs_cfg);
543 splinter_vlans = collect_splinter_vlans(ovs_cfg);
544 HMAP_FOR_EACH (br, node, &all_bridges) {
545 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
546 bridge_del_ports(br, &br->wanted_ports);
548 free(splinter_vlans);
550 /* Start pushing configuration changes down to the ofproto layer:
552 * - Delete ofprotos that are no longer configured.
554 * - Delete ports that are no longer configured.
556 * - Reconfigure existing ports to their desired configurations, or
557 * delete them if not possible.
559 * We have to do all the deletions before we can do any additions, because
560 * the ports to be added might require resources that will be freed up by
561 * deletions (they might especially overlap in name). */
562 bridge_delete_ofprotos();
563 HMAP_FOR_EACH (br, node, &all_bridges) {
565 bridge_delete_or_reconfigure_ports(br);
569 /* Finish pushing configuration changes to the ofproto layer:
571 * - Create ofprotos that are missing.
573 * - Add ports that are missing. */
574 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
578 error = ofproto_create(br->name, br->type, &br->ofproto);
580 VLOG_ERR("failed to create bridge %s: %s", br->name,
581 ovs_strerror(error));
582 shash_destroy(&br->wanted_ports);
587 HMAP_FOR_EACH (br, node, &all_bridges) {
588 bridge_add_ports(br, &br->wanted_ports);
589 shash_destroy(&br->wanted_ports);
592 reconfigure_system_stats(ovs_cfg);
594 /* Complete the configuration. */
595 sflow_bridge_number = 0;
596 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
597 HMAP_FOR_EACH (br, node, &all_bridges) {
600 /* We need the datapath ID early to allow LACP ports to use it as the
601 * default system ID. */
602 bridge_configure_datapath_id(br);
604 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
607 port_configure(port);
609 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
610 iface_set_ofport(iface->cfg, iface->ofp_port);
611 iface_configure_cfm(iface);
612 iface_configure_qos(iface, port->cfg->qos);
613 iface_set_mac(br, port, iface);
614 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
618 bridge_configure_mirrors(br);
619 bridge_configure_forward_bpdu(br);
620 bridge_configure_mac_table(br);
621 bridge_configure_remotes(br, managers, n_managers);
622 bridge_configure_netflow(br);
623 bridge_configure_sflow(br, &sflow_bridge_number);
624 bridge_configure_ipfix(br);
625 bridge_configure_stp(br);
626 bridge_configure_tables(br);
627 bridge_configure_dp_desc(br);
631 /* The ofproto-dpif provider does some final reconfiguration in its
632 * ->type_run() function. We have to call it before notifying the database
633 * client that reconfiguration is complete, otherwise there is a very
634 * narrow race window in which e.g. ofproto/trace will not recognize the
635 * new configuration (sometimes this causes unit test failures). */
639 /* Delete ofprotos which aren't configured or have the wrong type. Create
640 * ofprotos which don't exist but need to. */
642 bridge_delete_ofprotos(void)
649 /* Delete ofprotos with no bridge or with the wrong type. */
652 ofproto_enumerate_types(&types);
653 SSET_FOR_EACH (type, &types) {
656 ofproto_enumerate_names(type, &names);
657 SSET_FOR_EACH (name, &names) {
658 br = bridge_lookup(name);
659 if (!br || strcmp(type, br->type)) {
660 ofproto_delete(name, type);
664 sset_destroy(&names);
665 sset_destroy(&types);
669 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
671 if (*n >= *allocated) {
672 ports = x2nrealloc(ports, allocated, sizeof *ports);
674 ports[(*n)++] = port;
679 bridge_delete_or_reconfigure_ports(struct bridge *br)
681 struct ofproto_port ofproto_port;
682 struct ofproto_port_dump dump;
684 struct sset ofproto_ports;
685 struct port *port, *port_next;
687 /* List of "ofp_port"s to delete. We make a list instead of deleting them
688 * right away because ofproto implementations aren't necessarily able to
689 * iterate through a changing list of ports in an entirely robust way. */
696 sset_init(&ofproto_ports);
698 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
699 * that are not configured in the database. (This commonly happens when
700 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
702 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
703 * that have the wrong OpenFlow port number (and arrange to add them back
704 * with the correct OpenFlow port number). */
705 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
706 ofp_port_t requested_ofp_port;
709 sset_add(&ofproto_ports, ofproto_port.name);
711 iface = iface_lookup(br, ofproto_port.name);
713 /* No such iface is configured, so we should delete this
716 * As a corner case exception, keep the port if it's a bond fake
718 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
719 && !strcmp(ofproto_port.type, "internal")) {
725 if (strcmp(ofproto_port.type, iface->type)
726 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
727 /* The interface is the wrong type or can't be configured.
732 /* If the requested OpenFlow port for 'iface' changed, and it's not
733 * already the correct port, then we might want to temporarily delete
734 * this interface, so we can add it back again with the new OpenFlow
736 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
737 if (iface->ofp_port != OFPP_LOCAL &&
738 requested_ofp_port != OFPP_NONE &&
739 requested_ofp_port != iface->ofp_port) {
740 ofp_port_t victim_request;
741 struct iface *victim;
743 /* Check for an existing OpenFlow port currently occupying
744 * 'iface''s requested port number. If there isn't one, then
745 * delete this port. Otherwise we need to consider further. */
746 victim = iface_from_ofp_port(br, requested_ofp_port);
751 /* 'victim' is a port currently using 'iface''s requested port
752 * number. Unless 'victim' specifically requested that port
753 * number, too, then we can delete both 'iface' and 'victim'
754 * temporarily. (We'll add both of them back again later with new
755 * OpenFlow port numbers.)
757 * If 'victim' did request port number 'requested_ofp_port', just
758 * like 'iface', then that's a configuration inconsistency that we
759 * can't resolve. We might as well let it keep its current port
761 victim_request = iface_get_requested_ofp_port(victim->cfg);
762 if (victim_request != requested_ofp_port) {
763 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
764 iface_destroy(victim);
773 iface_destroy(iface);
774 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
776 for (i = 0; i < n; i++) {
777 ofproto_port_del(br->ofproto, del[i]);
781 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
782 * that we didn't see when we iterated through the datapath, i.e. ports
783 * that disappeared underneath use. This is an unusual situation, but it
784 * can happen in some cases:
786 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
787 * idea but could happen).
789 * - The port represented a device that disappeared, e.g. a tuntap
790 * device destroyed via "tunctl -d", a physical Ethernet device
791 * whose module was just unloaded via "rmmod", or a virtual NIC for a
792 * VM whose VM was just terminated. */
793 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
794 struct iface *iface, *iface_next;
796 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
797 if (!sset_contains(&ofproto_ports, iface->name)) {
798 iface_destroy__(iface);
802 if (list_is_empty(&port->ifaces)) {
806 sset_destroy(&ofproto_ports);
810 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
811 bool with_requested_port)
813 struct shash_node *port_node;
815 SHASH_FOR_EACH (port_node, wanted_ports) {
816 const struct ovsrec_port *port_cfg = port_node->data;
819 for (i = 0; i < port_cfg->n_interfaces; i++) {
820 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
821 ofp_port_t requested_ofp_port;
823 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
824 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
825 struct iface *iface = iface_lookup(br, iface_cfg->name);
828 iface_create(br, iface_cfg, port_cfg);
836 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
838 /* First add interfaces that request a particular port number. */
839 bridge_add_ports__(br, wanted_ports, true);
841 /* Then add interfaces that want automatic port number assignment.
842 * We add these afterward to avoid accidentally taking a specifically
843 * requested port number. */
844 bridge_add_ports__(br, wanted_ports, false);
848 port_configure(struct port *port)
850 const struct ovsrec_port *cfg = port->cfg;
851 struct bond_settings bond_settings;
852 struct lacp_settings lacp_settings;
853 struct ofproto_bundle_settings s;
856 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
857 configure_splinter_port(port);
866 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
867 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
868 s.slaves[s.n_slaves++] = iface->ofp_port;
873 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
877 /* Get VLAN trunks. */
880 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
884 if (cfg->vlan_mode) {
885 if (!strcmp(cfg->vlan_mode, "access")) {
886 s.vlan_mode = PORT_VLAN_ACCESS;
887 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
888 s.vlan_mode = PORT_VLAN_TRUNK;
889 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
890 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
891 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
892 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
894 /* This "can't happen" because ovsdb-server should prevent it. */
895 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
896 "back to trunk mode", port->name, cfg->vlan_mode);
897 s.vlan_mode = PORT_VLAN_TRUNK;
901 s.vlan_mode = PORT_VLAN_ACCESS;
903 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
907 s.vlan_mode = PORT_VLAN_TRUNK;
910 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
913 /* Get LACP settings. */
914 s.lacp = port_configure_lacp(port, &lacp_settings);
918 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
919 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
920 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
923 s.lacp_slaves = NULL;
926 /* Get bond settings. */
927 if (s.n_slaves > 1) {
928 s.bond = &bond_settings;
929 port_configure_bond(port, &bond_settings);
932 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
933 netdev_set_miimon_interval(iface->netdev, 0);
938 ofproto_bundle_register(port->bridge->ofproto, port, &s);
946 /* Pick local port hardware address and datapath ID for 'br'. */
948 bridge_configure_datapath_id(struct bridge *br)
950 uint8_t ea[ETH_ADDR_LEN];
952 struct iface *local_iface;
953 struct iface *hw_addr_iface;
956 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
957 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
959 int error = netdev_set_etheraddr(local_iface->netdev, ea);
961 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
962 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
963 "Ethernet address: %s",
964 br->name, ovs_strerror(error));
967 memcpy(br->ea, ea, ETH_ADDR_LEN);
969 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
970 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
971 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
972 ofproto_set_datapath_id(br->ofproto, dpid);
975 dpid_string = xasprintf("%016"PRIx64, dpid);
976 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
980 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
983 bridge_get_allowed_versions(struct bridge *br)
985 if (!br->cfg->n_protocols)
988 return ofputil_versions_from_strings(br->cfg->protocols,
989 br->cfg->n_protocols);
992 /* Set NetFlow configuration on 'br'. */
994 bridge_configure_netflow(struct bridge *br)
996 struct ovsrec_netflow *cfg = br->cfg->netflow;
997 struct netflow_options opts;
1000 ofproto_set_netflow(br->ofproto, NULL);
1004 memset(&opts, 0, sizeof opts);
1006 /* Get default NetFlow configuration from datapath.
1007 * Apply overrides from 'cfg'. */
1008 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1009 if (cfg->engine_type) {
1010 opts.engine_type = *cfg->engine_type;
1012 if (cfg->engine_id) {
1013 opts.engine_id = *cfg->engine_id;
1016 /* Configure active timeout interval. */
1017 opts.active_timeout = cfg->active_timeout;
1018 if (!opts.active_timeout) {
1019 opts.active_timeout = -1;
1020 } else if (opts.active_timeout < 0) {
1021 VLOG_WARN("bridge %s: active timeout interval set to negative "
1022 "value, using default instead (%d seconds)", br->name,
1023 NF_ACTIVE_TIMEOUT_DEFAULT);
1024 opts.active_timeout = -1;
1027 /* Add engine ID to interface number to disambiguate bridgs? */
1028 opts.add_id_to_iface = cfg->add_id_to_interface;
1029 if (opts.add_id_to_iface) {
1030 if (opts.engine_id > 0x7f) {
1031 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1032 "another vswitch, choose an engine id less than 128",
1035 if (hmap_count(&br->ports) > 508) {
1036 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1037 "another port when more than 508 ports are used",
1043 sset_init(&opts.collectors);
1044 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1047 if (ofproto_set_netflow(br->ofproto, &opts)) {
1048 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1050 sset_destroy(&opts.collectors);
1053 /* Set sFlow configuration on 'br'. */
1055 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1057 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1058 struct ovsrec_controller **controllers;
1059 struct ofproto_sflow_options oso;
1060 size_t n_controllers;
1064 ofproto_set_sflow(br->ofproto, NULL);
1068 memset(&oso, 0, sizeof oso);
1070 sset_init(&oso.targets);
1071 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1073 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1074 if (cfg->sampling) {
1075 oso.sampling_rate = *cfg->sampling;
1078 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1080 oso.polling_interval = *cfg->polling;
1083 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1085 oso.header_len = *cfg->header;
1088 oso.sub_id = (*sflow_bridge_number)++;
1089 oso.agent_device = cfg->agent;
1091 oso.control_ip = NULL;
1092 n_controllers = bridge_get_controllers(br, &controllers);
1093 for (i = 0; i < n_controllers; i++) {
1094 if (controllers[i]->local_ip) {
1095 oso.control_ip = controllers[i]->local_ip;
1099 ofproto_set_sflow(br->ofproto, &oso);
1101 sset_destroy(&oso.targets);
1104 /* Returns whether a IPFIX row is valid. */
1106 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1108 return ipfix && ipfix->n_targets > 0;
1111 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1113 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1114 const struct bridge *br)
1116 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1119 /* Set IPFIX configuration on 'br'. */
1121 bridge_configure_ipfix(struct bridge *br)
1123 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1124 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1125 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1126 struct ofproto_ipfix_bridge_exporter_options be_opts;
1127 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1128 size_t n_fe_opts = 0;
1130 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1131 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1136 if (!valid_be_cfg && n_fe_opts == 0) {
1137 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1142 memset(&be_opts, 0, sizeof be_opts);
1144 sset_init(&be_opts.targets);
1145 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1147 if (be_cfg->sampling) {
1148 be_opts.sampling_rate = *be_cfg->sampling;
1150 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1152 if (be_cfg->obs_domain_id) {
1153 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1155 if (be_cfg->obs_point_id) {
1156 be_opts.obs_point_id = *be_cfg->obs_point_id;
1158 if (be_cfg->cache_active_timeout) {
1159 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1161 if (be_cfg->cache_max_flows) {
1162 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1166 if (n_fe_opts > 0) {
1167 struct ofproto_ipfix_flow_exporter_options *opts;
1168 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1170 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1171 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1172 opts->collector_set_id = fe_cfg->id;
1173 sset_init(&opts->targets);
1174 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1175 fe_cfg->ipfix->n_targets);
1176 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1177 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1178 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1179 ? *fe_cfg->ipfix->cache_max_flows : 0;
1185 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1189 sset_destroy(&be_opts.targets);
1192 if (n_fe_opts > 0) {
1193 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1195 for (i = 0; i < n_fe_opts; i++) {
1196 sset_destroy(&opts->targets);
1204 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1205 struct ofproto_port_stp_settings *port_s,
1206 int *port_num_counter, unsigned long *port_num_bitmap)
1208 const char *config_str;
1209 struct iface *iface;
1211 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1212 port_s->enable = false;
1215 port_s->enable = true;
1218 /* STP over bonds is not supported. */
1219 if (!list_is_singleton(&port->ifaces)) {
1220 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1222 port_s->enable = false;
1226 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1228 /* Internal ports shouldn't participate in spanning tree, so
1230 if (!strcmp(iface->type, "internal")) {
1231 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1232 port_s->enable = false;
1236 /* STP on mirror output ports is not supported. */
1237 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1238 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1239 port_s->enable = false;
1243 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1245 unsigned long int port_num = strtoul(config_str, NULL, 0);
1246 int port_idx = port_num - 1;
1248 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1249 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1250 port_s->enable = false;
1254 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1255 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1256 port->name, port_num);
1257 port_s->enable = false;
1260 bitmap_set1(port_num_bitmap, port_idx);
1261 port_s->port_num = port_idx;
1263 if (*port_num_counter >= STP_MAX_PORTS) {
1264 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1265 port_s->enable = false;
1269 port_s->port_num = (*port_num_counter)++;
1272 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1274 port_s->path_cost = strtoul(config_str, NULL, 10);
1276 enum netdev_features current;
1279 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1280 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1281 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1284 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1286 port_s->priority = strtoul(config_str, NULL, 0);
1288 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1292 /* Set spanning tree configuration on 'br'. */
1294 bridge_configure_stp(struct bridge *br)
1296 if (!br->cfg->stp_enable) {
1297 ofproto_set_stp(br->ofproto, NULL);
1299 struct ofproto_stp_settings br_s;
1300 const char *config_str;
1302 int port_num_counter;
1303 unsigned long *port_num_bitmap;
1305 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1307 uint8_t ea[ETH_ADDR_LEN];
1309 if (eth_addr_from_string(config_str, ea)) {
1310 br_s.system_id = eth_addr_to_uint64(ea);
1312 br_s.system_id = eth_addr_to_uint64(br->ea);
1313 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1314 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1317 br_s.system_id = eth_addr_to_uint64(br->ea);
1320 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1322 br_s.priority = strtoul(config_str, NULL, 0);
1324 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1327 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1329 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1331 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1334 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1336 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1338 br_s.max_age = STP_DEFAULT_MAX_AGE;
1341 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1343 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1345 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1348 /* Configure STP on the bridge. */
1349 if (ofproto_set_stp(br->ofproto, &br_s)) {
1350 VLOG_ERR("bridge %s: could not enable STP", br->name);
1354 /* Users must either set the port number with the "stp-port-num"
1355 * configuration on all ports or none. If manual configuration
1356 * is not done, then we allocate them sequentially. */
1357 port_num_counter = 0;
1358 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1359 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1360 struct ofproto_port_stp_settings port_s;
1361 struct iface *iface;
1363 port_configure_stp(br->ofproto, port, &port_s,
1364 &port_num_counter, port_num_bitmap);
1366 /* As bonds are not supported, just apply configuration to
1367 * all interfaces. */
1368 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1369 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1371 VLOG_ERR("port %s: could not enable STP", port->name);
1377 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1378 && port_num_counter) {
1379 VLOG_ERR("bridge %s: must manually configure all STP port "
1380 "IDs or none, disabling", br->name);
1381 ofproto_set_stp(br->ofproto, NULL);
1383 bitmap_free(port_num_bitmap);
1388 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1390 const struct port *port = port_lookup(br, name);
1391 return port && port_is_bond_fake_iface(port);
1395 port_is_bond_fake_iface(const struct port *port)
1397 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1401 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1403 struct bridge *br, *next;
1404 struct shash new_br;
1407 /* Collect new bridges' names and types. */
1408 shash_init(&new_br);
1409 for (i = 0; i < cfg->n_bridges; i++) {
1410 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1411 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1413 if (strchr(br_cfg->name, '/')) {
1414 /* Prevent remote ovsdb-server users from accessing arbitrary
1415 * directories, e.g. consider a bridge named "../../../etc/". */
1416 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1418 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1419 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1423 /* Get rid of deleted bridges or those whose types have changed.
1424 * Update 'cfg' of bridges that still exist. */
1425 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1426 br->cfg = shash_find_data(&new_br, br->name);
1427 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1428 br->cfg->datapath_type))) {
1433 /* Add new bridges. */
1434 for (i = 0; i < cfg->n_bridges; i++) {
1435 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1436 struct bridge *br = bridge_lookup(br_cfg->name);
1438 bridge_create(br_cfg);
1442 shash_destroy(&new_br);
1445 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1446 * Returns 0 if successful, otherwise a positive errno value. */
1448 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1449 struct netdev *netdev)
1451 return netdev_set_config(netdev, &iface_cfg->options);
1454 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1455 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1457 * If successful, returns 0 and stores the network device in '*netdevp'. On
1458 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1460 iface_do_create(const struct bridge *br,
1461 const struct ovsrec_interface *iface_cfg,
1462 const struct ovsrec_port *port_cfg,
1463 ofp_port_t *ofp_portp, struct netdev **netdevp)
1465 struct netdev *netdev = NULL;
1468 if (netdev_is_reserved_name(iface_cfg->name)) {
1469 VLOG_WARN("could not create interface %s, name is reserved",
1475 error = netdev_open(iface_cfg->name,
1476 iface_get_type(iface_cfg, br->cfg), &netdev);
1478 VLOG_WARN("could not open network device %s (%s)",
1479 iface_cfg->name, ovs_strerror(error));
1483 error = iface_set_netdev_config(iface_cfg, netdev);
1488 *ofp_portp = iface_pick_ofport(iface_cfg);
1489 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1494 VLOG_INFO("bridge %s: added interface %s on port %d",
1495 br->name, iface_cfg->name, *ofp_portp);
1497 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1498 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1506 netdev_close(netdev);
1510 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1511 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1512 * automatically allocated for the iface. Takes ownership of and
1513 * deallocates 'if_cfg'.
1515 * Return true if an iface is successfully created, false otherwise. */
1517 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1518 const struct ovsrec_port *port_cfg)
1520 struct netdev *netdev;
1521 struct iface *iface;
1522 ofp_port_t ofp_port;
1526 /* Do the bits that can fail up front. */
1527 ovs_assert(!iface_lookup(br, iface_cfg->name));
1528 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1530 iface_clear_db_record(iface_cfg);
1534 /* Get or create the port structure. */
1535 port = port_lookup(br, port_cfg->name);
1537 port = port_create(br, port_cfg);
1540 /* Create the iface structure. */
1541 iface = xzalloc(sizeof *iface);
1542 list_push_back(&port->ifaces, &iface->port_elem);
1543 hmap_insert(&br->iface_by_name, &iface->name_node,
1544 hash_string(iface_cfg->name, 0));
1546 iface->name = xstrdup(iface_cfg->name);
1547 iface->ofp_port = ofp_port;
1548 iface->netdev = netdev;
1549 iface->type = iface_get_type(iface_cfg, br->cfg);
1550 iface->cfg = iface_cfg;
1551 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1552 hash_ofp_port(ofp_port));
1554 /* Populate initial status in database. */
1555 iface_refresh_stats(iface);
1557 /* Add bond fake iface if necessary. */
1558 if (port_is_bond_fake_iface(port)) {
1559 struct ofproto_port ofproto_port;
1561 if (ofproto_port_query_by_name(br->ofproto, port->name,
1563 struct netdev *netdev;
1566 error = netdev_open(port->name, "internal", &netdev);
1568 ofp_port_t fake_ofp_port = OFPP_NONE;
1569 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1570 netdev_close(netdev);
1572 VLOG_WARN("could not open network device %s (%s)",
1573 port->name, ovs_strerror(error));
1576 /* Already exists, nothing to do. */
1577 ofproto_port_destroy(&ofproto_port);
1584 /* Set forward BPDU option. */
1586 bridge_configure_forward_bpdu(struct bridge *br)
1588 ofproto_set_forward_bpdu(br->ofproto,
1589 smap_get_bool(&br->cfg->other_config,
1594 /* Set MAC learning table configuration for 'br'. */
1596 bridge_configure_mac_table(struct bridge *br)
1598 const char *idle_time_str;
1601 const char *mac_table_size_str;
1604 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1605 idle_time = (idle_time_str && atoi(idle_time_str)
1606 ? atoi(idle_time_str)
1607 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1609 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1610 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1611 ? atoi(mac_table_size_str)
1614 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1618 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1619 const struct port *fake_br, struct iface **hw_addr_iface)
1621 struct hmapx mirror_output_ports;
1623 bool found_addr = false;
1627 /* Mirror output ports don't participate in picking the local hardware
1628 * address. ofproto can't help us find out whether a given port is a
1629 * mirror output because we haven't configured mirrors yet, so we need to
1630 * accumulate them ourselves. */
1631 hmapx_init(&mirror_output_ports);
1632 for (i = 0; i < br->cfg->n_mirrors; i++) {
1633 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1634 if (m->output_port) {
1635 hmapx_add(&mirror_output_ports, m->output_port);
1639 /* Otherwise choose the minimum non-local MAC address among all of the
1641 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1642 uint8_t iface_ea[ETH_ADDR_LEN];
1643 struct iface *candidate;
1644 struct iface *iface;
1646 /* Mirror output ports don't participate. */
1647 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1651 /* Choose the MAC address to represent the port. */
1653 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1654 /* Find the interface with this Ethernet address (if any) so that
1655 * we can provide the correct devname to the caller. */
1656 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1657 uint8_t candidate_ea[ETH_ADDR_LEN];
1658 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1659 && eth_addr_equals(iface_ea, candidate_ea)) {
1664 /* Choose the interface whose MAC address will represent the port.
1665 * The Linux kernel bonding code always chooses the MAC address of
1666 * the first slave added to a bond, and the Fedora networking
1667 * scripts always add slaves to a bond in alphabetical order, so
1668 * for compatibility we choose the interface with the name that is
1669 * first in alphabetical order. */
1670 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1671 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1676 /* The local port doesn't count (since we're trying to choose its
1677 * MAC address anyway). */
1678 if (iface->ofp_port == OFPP_LOCAL) {
1682 /* For fake bridges we only choose from ports with the same tag */
1683 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1684 if (!port->cfg->tag) {
1687 if (*port->cfg->tag != *fake_br->cfg->tag) {
1693 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1699 /* Compare against our current choice. */
1700 if (!eth_addr_is_multicast(iface_ea) &&
1701 !eth_addr_is_local(iface_ea) &&
1702 !eth_addr_is_reserved(iface_ea) &&
1703 !eth_addr_is_zero(iface_ea) &&
1704 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1706 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1707 *hw_addr_iface = iface;
1713 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1714 *hw_addr_iface = NULL;
1717 hmapx_destroy(&mirror_output_ports);
1721 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1722 struct iface **hw_addr_iface)
1725 *hw_addr_iface = NULL;
1727 /* Did the user request a particular MAC? */
1728 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1729 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1730 if (eth_addr_is_multicast(ea)) {
1731 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1732 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1733 } else if (eth_addr_is_zero(ea)) {
1734 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1740 /* Find a local hw address */
1741 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1744 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1745 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1746 * an interface on 'br', then that interface must be passed in as
1747 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1748 * 'hw_addr_iface' must be passed in as a null pointer. */
1750 bridge_pick_datapath_id(struct bridge *br,
1751 const uint8_t bridge_ea[ETH_ADDR_LEN],
1752 struct iface *hw_addr_iface)
1755 * The procedure for choosing a bridge MAC address will, in the most
1756 * ordinary case, also choose a unique MAC that we can use as a datapath
1757 * ID. In some special cases, though, multiple bridges will end up with
1758 * the same MAC address. This is OK for the bridges, but it will confuse
1759 * the OpenFlow controller, because each datapath needs a unique datapath
1762 * Datapath IDs must be unique. It is also very desirable that they be
1763 * stable from one run to the next, so that policy set on a datapath
1766 const char *datapath_id;
1769 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1770 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1774 if (!hw_addr_iface) {
1776 * A purely internal bridge, that is, one that has no non-virtual
1777 * network devices on it at all, is difficult because it has no
1778 * natural unique identifier at all.
1780 * When the host is a XenServer, we handle this case by hashing the
1781 * host's UUID with the name of the bridge. Names of bridges are
1782 * persistent across XenServer reboots, although they can be reused if
1783 * an internal network is destroyed and then a new one is later
1784 * created, so this is fairly effective.
1786 * When the host is not a XenServer, we punt by using a random MAC
1787 * address on each run.
1789 const char *host_uuid = xenserver_get_host_uuid();
1791 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1792 dpid = dpid_from_hash(combined, strlen(combined));
1798 return eth_addr_to_uint64(bridge_ea);
1802 dpid_from_hash(const void *data, size_t n)
1804 uint8_t hash[SHA1_DIGEST_SIZE];
1806 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1807 sha1_bytes(data, n, hash);
1808 eth_addr_mark_random(hash);
1809 return eth_addr_to_uint64(hash);
1813 iface_refresh_netdev_status(struct iface *iface)
1817 enum netdev_features current;
1818 enum netdev_flags flags;
1819 const char *link_state;
1820 uint8_t mac[ETH_ADDR_LEN];
1821 int64_t bps, mtu_64, ifindex64, link_resets;
1824 if (iface_is_synthetic(iface)) {
1828 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
1829 && !force_status_commit) {
1833 iface->change_seq = netdev_get_change_seq(iface->netdev);
1837 if (!netdev_get_status(iface->netdev, &smap)) {
1838 ovsrec_interface_set_status(iface->cfg, &smap);
1840 ovsrec_interface_set_status(iface->cfg, NULL);
1843 smap_destroy(&smap);
1845 error = netdev_get_flags(iface->netdev, &flags);
1847 const char *state = flags & NETDEV_UP ? "up" : "down";
1849 ovsrec_interface_set_admin_state(iface->cfg, state);
1851 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1854 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1855 ovsrec_interface_set_link_state(iface->cfg, link_state);
1857 link_resets = netdev_get_carrier_resets(iface->netdev);
1858 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1860 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1861 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1863 ovsrec_interface_set_duplex(iface->cfg,
1864 netdev_features_is_full_duplex(current)
1866 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1868 ovsrec_interface_set_duplex(iface->cfg, NULL);
1869 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1872 error = netdev_get_mtu(iface->netdev, &mtu);
1875 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1877 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1880 error = netdev_get_etheraddr(iface->netdev, mac);
1882 char mac_string[32];
1884 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1885 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1887 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1890 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1891 * if there is no valid ifindex number. */
1892 ifindex64 = netdev_get_ifindex(iface->netdev);
1893 if (ifindex64 < 0) {
1896 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1900 iface_refresh_ofproto_status(struct iface *iface)
1905 if (iface_is_synthetic(iface)) {
1909 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1913 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1915 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1918 iface_refresh_cfm_stats(iface);
1921 error = ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1922 iface->ofp_port, force_status_commit,
1925 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1927 smap_destroy(&smap);
1930 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1933 iface_refresh_cfm_stats(struct iface *iface)
1935 const struct ovsrec_interface *cfg = iface->cfg;
1936 struct ofproto_cfm_status status;
1939 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1940 iface->ofp_port, force_status_commit,
1943 /* Do nothing if there is no status change since last update. */
1944 } else if (error > 0) {
1945 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1946 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1947 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1948 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1949 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1950 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1952 const char *reasons[CFM_FAULT_N_REASONS];
1953 int64_t cfm_health = status.health;
1954 int64_t cfm_flap_count = status.flap_count;
1955 bool faulted = status.faults != 0;
1958 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1961 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1962 int reason = 1 << i;
1963 if (status.faults & reason) {
1964 reasons[j++] = cfm_fault_reason_to_str(reason);
1967 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1969 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1971 if (status.remote_opstate >= 0) {
1972 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1973 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1975 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1978 ovsrec_interface_set_cfm_remote_mpids(cfg,
1979 (const int64_t *)status.rmps,
1981 if (cfm_health >= 0) {
1982 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1984 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1992 iface_refresh_stats(struct iface *iface)
1994 #define IFACE_STATS \
1995 IFACE_STAT(rx_packets, "rx_packets") \
1996 IFACE_STAT(tx_packets, "tx_packets") \
1997 IFACE_STAT(rx_bytes, "rx_bytes") \
1998 IFACE_STAT(tx_bytes, "tx_bytes") \
1999 IFACE_STAT(rx_dropped, "rx_dropped") \
2000 IFACE_STAT(tx_dropped, "tx_dropped") \
2001 IFACE_STAT(rx_errors, "rx_errors") \
2002 IFACE_STAT(tx_errors, "tx_errors") \
2003 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2004 IFACE_STAT(rx_over_errors, "rx_over_err") \
2005 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2006 IFACE_STAT(collisions, "collisions")
2008 #define IFACE_STAT(MEMBER, NAME) + 1
2009 enum { N_IFACE_STATS = IFACE_STATS };
2011 int64_t values[N_IFACE_STATS];
2012 char *keys[N_IFACE_STATS];
2015 struct netdev_stats stats;
2017 if (iface_is_synthetic(iface)) {
2021 /* Intentionally ignore return value, since errors will set 'stats' to
2022 * all-1s, and we will deal with that correctly below. */
2023 netdev_get_stats(iface->netdev, &stats);
2025 /* Copy statistics into keys[] and values[]. */
2027 #define IFACE_STAT(MEMBER, NAME) \
2028 if (stats.MEMBER != UINT64_MAX) { \
2030 values[n] = stats.MEMBER; \
2035 ovs_assert(n <= N_IFACE_STATS);
2037 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2042 br_refresh_stp_status(struct bridge *br)
2044 struct smap smap = SMAP_INITIALIZER(&smap);
2045 struct ofproto *ofproto = br->ofproto;
2046 struct ofproto_stp_status status;
2048 if (ofproto_get_stp_status(ofproto, &status)) {
2052 if (!status.enabled) {
2053 ovsrec_bridge_set_status(br->cfg, NULL);
2057 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2058 STP_ID_ARGS(status.bridge_id));
2059 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2060 STP_ID_ARGS(status.designated_root));
2061 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2063 ovsrec_bridge_set_status(br->cfg, &smap);
2064 smap_destroy(&smap);
2068 port_refresh_stp_status(struct port *port)
2070 struct ofproto *ofproto = port->bridge->ofproto;
2071 struct iface *iface;
2072 struct ofproto_port_stp_status status;
2075 if (port_is_synthetic(port)) {
2079 /* STP doesn't currently support bonds. */
2080 if (!list_is_singleton(&port->ifaces)) {
2081 ovsrec_port_set_status(port->cfg, NULL);
2085 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2086 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2090 if (!status.enabled) {
2091 ovsrec_port_set_status(port->cfg, NULL);
2095 /* Set Status column. */
2097 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2098 smap_add(&smap, "stp_state", stp_state_name(status.state));
2099 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2100 smap_add(&smap, "stp_role", stp_role_name(status.role));
2101 ovsrec_port_set_status(port->cfg, &smap);
2102 smap_destroy(&smap);
2106 port_refresh_stp_stats(struct port *port)
2108 struct ofproto *ofproto = port->bridge->ofproto;
2109 struct iface *iface;
2110 struct ofproto_port_stp_stats stats;
2112 int64_t int_values[3];
2114 if (port_is_synthetic(port)) {
2118 /* STP doesn't currently support bonds. */
2119 if (!list_is_singleton(&port->ifaces)) {
2123 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2124 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2128 if (!stats.enabled) {
2129 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2133 /* Set Statistics column. */
2134 keys[0] = "stp_tx_count";
2135 int_values[0] = stats.tx_count;
2136 keys[1] = "stp_rx_count";
2137 int_values[1] = stats.rx_count;
2138 keys[2] = "stp_error_count";
2139 int_values[2] = stats.error_count;
2141 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2142 ARRAY_SIZE(int_values));
2146 port_refresh_bond_status(struct port *port, bool force_update)
2150 /* Return if port is not a bond */
2151 if (list_is_singleton(&port->ifaces)) {
2155 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2159 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2160 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2166 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2168 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2172 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2174 bool enable = enable_system_stats(cfg);
2176 system_stats_enable(enable);
2178 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2183 run_system_stats(void)
2185 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2188 stats = system_stats_run();
2190 struct ovsdb_idl_txn *txn;
2191 struct ovsdb_datum datum;
2193 txn = ovsdb_idl_txn_create(idl);
2194 ovsdb_datum_from_smap(&datum, stats);
2195 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2197 ovsdb_idl_txn_commit(txn);
2198 ovsdb_idl_txn_destroy(txn);
2205 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2208 case OFPCR12_ROLE_EQUAL:
2210 case OFPCR12_ROLE_MASTER:
2212 case OFPCR12_ROLE_SLAVE:
2214 case OFPCR12_ROLE_NOCHANGE:
2216 return "*** INVALID ROLE ***";
2221 refresh_controller_status(void)
2225 const struct ovsrec_controller *cfg;
2229 /* Accumulate status for controllers on all bridges. */
2230 HMAP_FOR_EACH (br, node, &all_bridges) {
2231 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2234 /* Update each controller in the database with current status. */
2235 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2236 struct ofproto_controller_info *cinfo =
2237 shash_find_data(&info, cfg->target);
2240 struct smap smap = SMAP_INITIALIZER(&smap);
2241 const char **values = cinfo->pairs.values;
2242 const char **keys = cinfo->pairs.keys;
2245 for (i = 0; i < cinfo->pairs.n; i++) {
2246 smap_add(&smap, keys[i], values[i]);
2249 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2250 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2252 ovsrec_controller_set_status(cfg, &smap);
2253 smap_destroy(&smap);
2255 ovsrec_controller_set_is_connected(cfg, false);
2256 ovsrec_controller_set_role(cfg, NULL);
2257 ovsrec_controller_set_status(cfg, NULL);
2261 ofproto_free_ofproto_controller_info(&info);
2271 /* Let each datapath type do the work that it needs to do. */
2273 ofproto_enumerate_types(&types);
2274 SSET_FOR_EACH (type, &types) {
2275 ofproto_type_run(type);
2277 sset_destroy(&types);
2279 /* Let each bridge do the work that it needs to do. */
2280 HMAP_FOR_EACH (br, node, &all_bridges) {
2281 ofproto_run(br->ofproto);
2288 static struct ovsrec_open_vswitch null_cfg;
2289 const struct ovsrec_open_vswitch *cfg;
2291 bool vlan_splinters_changed;
2295 ovsrec_open_vswitch_init(&null_cfg);
2299 if (ovsdb_idl_is_lock_contended(idl)) {
2300 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2301 struct bridge *br, *next_br;
2303 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2304 "disabling this process (pid %ld) until it goes away",
2305 (long int) getpid());
2307 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2310 /* Since we will not be running system_stats_run() in this process
2311 * with the current situation of multiple ovs-vswitchd daemons,
2312 * disable system stats collection. */
2313 system_stats_enable(false);
2314 /* This prevents the process from constantly waking up on
2315 * connectivity seq. */
2316 connectivity_seqno = seq_read(connectivity_seq_get());
2318 } else if (!ovsdb_idl_has_lock(idl)) {
2321 cfg = ovsrec_open_vswitch_first(idl);
2323 /* Initialize the ofproto library. This only needs to run once, but
2324 * it must be done after the configuration is set. If the
2325 * initialization has already occurred, bridge_init_ofproto()
2326 * returns immediately. */
2327 bridge_init_ofproto(cfg);
2329 /* Once the value of flow-restore-wait is false, we no longer should
2330 * check its value from the database. */
2331 if (cfg && ofproto_get_flow_restore_wait()) {
2332 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2333 "flow-restore-wait", false));
2338 /* Re-configure SSL. We do this on every trip through the main loop,
2339 * instead of just when the database changes, because the contents of the
2340 * key and certificate files can change without the database changing.
2342 * We do this before bridge_reconfigure() because that function might
2343 * initiate SSL connections and thus requires SSL to be configured. */
2344 if (cfg && cfg->ssl) {
2345 const struct ovsrec_ssl *ssl = cfg->ssl;
2347 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2348 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2351 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2352 * usage has changed. */
2353 vlan_splinters_changed = false;
2354 if (vlan_splinters_enabled_anywhere) {
2355 HMAP_FOR_EACH (br, node, &all_bridges) {
2356 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2357 vlan_splinters_changed = true;
2363 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2364 struct ovsdb_idl_txn *txn;
2366 idl_seqno = ovsdb_idl_get_seqno(idl);
2367 txn = ovsdb_idl_txn_create(idl);
2368 bridge_reconfigure(cfg ? cfg : &null_cfg);
2371 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2374 /* If we are completing our initial configuration for this run
2375 * of ovs-vswitchd, then keep the transaction around to monitor
2376 * it for completion. */
2377 if (initial_config_done) {
2378 ovsdb_idl_txn_commit(txn);
2379 ovsdb_idl_txn_destroy(txn);
2381 initial_config_done = true;
2382 daemonize_txn = txn;
2386 if (daemonize_txn) {
2387 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2388 if (status != TXN_INCOMPLETE) {
2389 ovsdb_idl_txn_destroy(daemonize_txn);
2390 daemonize_txn = NULL;
2392 /* ovs-vswitchd has completed initialization, so allow the
2393 * process that forked us to exit successfully. */
2394 daemonize_complete();
2396 vlog_enable_async();
2398 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2402 /* Statistics update interval should always be greater than or equal to
2405 stats_interval = MAX(smap_get_int(&cfg->other_config,
2406 "stats-update-interval",
2409 stats_interval = 5000;
2411 if (stats_timer_interval != stats_interval) {
2412 stats_timer_interval = stats_interval;
2413 stats_timer = LLONG_MIN;
2416 /* Refresh interface and mirror stats if necessary. */
2417 if (time_msec() >= stats_timer && cfg) {
2418 enum ovsdb_idl_txn_status status;
2420 /* Rate limit the update. Do not start a new update if the
2421 * previous one is not done. */
2423 stats_txn = ovsdb_idl_txn_create(idl);
2424 HMAP_FOR_EACH (br, node, &all_bridges) {
2428 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2429 struct iface *iface;
2431 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2432 iface_refresh_stats(iface);
2434 port_refresh_stp_stats(port);
2436 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2437 mirror_refresh_stats(m);
2440 refresh_controller_status();
2443 status = ovsdb_idl_txn_commit(stats_txn);
2444 if (status != TXN_INCOMPLETE) {
2445 stats_timer = time_msec() + stats_timer_interval;
2446 ovsdb_idl_txn_destroy(stats_txn);
2454 /* Check the need to update status. */
2455 seq = seq_read(connectivity_seq_get());
2456 if (seq != connectivity_seqno || force_status_commit) {
2457 connectivity_seqno = seq;
2458 status_txn = ovsdb_idl_txn_create(idl);
2459 HMAP_FOR_EACH (br, node, &all_bridges) {
2462 br_refresh_stp_status(br);
2463 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2464 struct iface *iface;
2466 port_refresh_stp_status(port);
2467 port_refresh_bond_status(port, force_status_commit);
2468 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2469 iface_refresh_netdev_status(iface);
2470 iface_refresh_ofproto_status(iface);
2478 enum ovsdb_idl_txn_status status;
2480 status = ovsdb_idl_txn_commit(status_txn);
2482 /* If the transaction is incomplete or fails, 'status_txn'
2483 * needs to be committed next iteration of bridge_run() even if
2484 * connectivity or netdev sequence numbers do not change. */
2485 if (status == TXN_SUCCESS || status == TXN_UNCHANGED)
2487 force_status_commit = false;
2489 force_status_commit = true;
2492 /* Do not destroy "status_txn" if the transaction is
2493 * "TXN_INCOMPLETE". */
2494 if (status != TXN_INCOMPLETE) {
2495 ovsdb_idl_txn_destroy(status_txn);
2509 ovsdb_idl_wait(idl);
2510 if (daemonize_txn) {
2511 ovsdb_idl_txn_wait(daemonize_txn);
2515 ofproto_enumerate_types(&types);
2516 SSET_FOR_EACH (type, &types) {
2517 ofproto_type_wait(type);
2519 sset_destroy(&types);
2521 if (!hmap_is_empty(&all_bridges)) {
2524 HMAP_FOR_EACH (br, node, &all_bridges) {
2525 ofproto_wait(br->ofproto);
2528 poll_timer_wait_until(stats_timer);
2531 /* If the status database transaction is 'TXN_INCOMPLETE' or is
2532 * unsuccessful, register a timeout in 'STATUS_CHECK_AGAIN_MSEC'. Else,
2533 * wait on the global connectivity sequence number. Note, this also helps
2534 * batch multiple status changes into one transaction. */
2535 if (force_status_commit) {
2536 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2538 seq_wait(connectivity_seq_get(), connectivity_seqno);
2541 system_stats_wait();
2544 /* Adds some memory usage statistics for bridges into 'usage', for use with
2545 * memory_report(). */
2547 bridge_get_memory_usage(struct simap *usage)
2554 ofproto_enumerate_types(&types);
2555 SSET_FOR_EACH (type, &types) {
2556 ofproto_type_get_memory_usage(type, usage);
2558 sset_destroy(&types);
2560 HMAP_FOR_EACH (br, node, &all_bridges) {
2561 ofproto_get_memory_usage(br->ofproto, usage);
2565 /* QoS unixctl user interface functions. */
2567 struct qos_unixctl_show_cbdata {
2569 struct iface *iface;
2573 qos_unixctl_show_queue(unsigned int queue_id,
2574 const struct smap *details,
2575 struct iface *iface,
2578 struct netdev_queue_stats stats;
2579 struct smap_node *node;
2582 ds_put_cstr(ds, "\n");
2584 ds_put_format(ds, "Queue %u:\n", queue_id);
2586 ds_put_cstr(ds, "Default:\n");
2589 SMAP_FOR_EACH (node, details) {
2590 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2593 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2595 if (stats.tx_packets != UINT64_MAX) {
2596 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2599 if (stats.tx_bytes != UINT64_MAX) {
2600 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2603 if (stats.tx_errors != UINT64_MAX) {
2604 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2607 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2608 queue_id, ovs_strerror(error));
2613 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2614 const char *argv[], void *aux OVS_UNUSED)
2616 struct ds ds = DS_EMPTY_INITIALIZER;
2617 struct smap smap = SMAP_INITIALIZER(&smap);
2618 struct iface *iface;
2620 struct smap_node *node;
2622 iface = iface_find(argv[1]);
2624 unixctl_command_reply_error(conn, "no such interface");
2628 netdev_get_qos(iface->netdev, &type, &smap);
2630 if (*type != '\0') {
2631 struct netdev_queue_dump dump;
2632 struct smap details;
2633 unsigned int queue_id;
2635 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2637 SMAP_FOR_EACH (node, &smap) {
2638 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2641 smap_init(&details);
2642 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2643 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2645 smap_destroy(&details);
2647 unixctl_command_reply(conn, ds_cstr(&ds));
2649 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2650 unixctl_command_reply_error(conn, ds_cstr(&ds));
2653 smap_destroy(&smap);
2657 /* Bridge reconfiguration functions. */
2659 bridge_create(const struct ovsrec_bridge *br_cfg)
2663 ovs_assert(!bridge_lookup(br_cfg->name));
2664 br = xzalloc(sizeof *br);
2666 br->name = xstrdup(br_cfg->name);
2667 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2670 /* Derive the default Ethernet address from the bridge's UUID. This should
2671 * be unique and it will be stable between ovs-vswitchd runs. */
2672 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2673 eth_addr_mark_random(br->default_ea);
2675 hmap_init(&br->ports);
2676 hmap_init(&br->ifaces);
2677 hmap_init(&br->iface_by_name);
2678 hmap_init(&br->mirrors);
2680 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2684 bridge_destroy(struct bridge *br)
2687 struct mirror *mirror, *next_mirror;
2688 struct port *port, *next_port;
2690 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2693 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2694 mirror_destroy(mirror);
2697 hmap_remove(&all_bridges, &br->node);
2698 ofproto_destroy(br->ofproto);
2699 hmap_destroy(&br->ifaces);
2700 hmap_destroy(&br->ports);
2701 hmap_destroy(&br->iface_by_name);
2702 hmap_destroy(&br->mirrors);
2709 static struct bridge *
2710 bridge_lookup(const char *name)
2714 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2715 if (!strcmp(br->name, name)) {
2722 /* Handle requests for a listing of all flows known by the OpenFlow
2723 * stack, including those normally hidden. */
2725 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2726 const char *argv[], void *aux OVS_UNUSED)
2731 br = bridge_lookup(argv[1]);
2733 unixctl_command_reply_error(conn, "Unknown bridge");
2738 ofproto_get_all_flows(br->ofproto, &results);
2740 unixctl_command_reply(conn, ds_cstr(&results));
2741 ds_destroy(&results);
2744 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2745 * connections and reconnect. If BRIDGE is not specified, then all bridges
2746 * drop their controller connections and reconnect. */
2748 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2749 const char *argv[], void *aux OVS_UNUSED)
2753 br = bridge_lookup(argv[1]);
2755 unixctl_command_reply_error(conn, "Unknown bridge");
2758 ofproto_reconnect_controllers(br->ofproto);
2760 HMAP_FOR_EACH (br, node, &all_bridges) {
2761 ofproto_reconnect_controllers(br->ofproto);
2764 unixctl_command_reply(conn, NULL);
2768 bridge_get_controllers(const struct bridge *br,
2769 struct ovsrec_controller ***controllersp)
2771 struct ovsrec_controller **controllers;
2772 size_t n_controllers;
2774 controllers = br->cfg->controller;
2775 n_controllers = br->cfg->n_controller;
2777 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2783 *controllersp = controllers;
2785 return n_controllers;
2789 bridge_collect_wanted_ports(struct bridge *br,
2790 const unsigned long int *splinter_vlans,
2791 struct shash *wanted_ports)
2795 shash_init(wanted_ports);
2797 for (i = 0; i < br->cfg->n_ports; i++) {
2798 const char *name = br->cfg->ports[i]->name;
2799 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2800 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2804 if (bridge_get_controllers(br, NULL)
2805 && !shash_find(wanted_ports, br->name)) {
2806 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2807 br->name, br->name);
2809 ovsrec_interface_init(&br->synth_local_iface);
2810 ovsrec_port_init(&br->synth_local_port);
2812 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2813 br->synth_local_port.n_interfaces = 1;
2814 br->synth_local_port.name = br->name;
2816 br->synth_local_iface.name = br->name;
2817 br->synth_local_iface.type = "internal";
2819 br->synth_local_ifacep = &br->synth_local_iface;
2821 shash_add(wanted_ports, br->name, &br->synth_local_port);
2824 if (splinter_vlans) {
2825 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2829 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2830 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2831 * 'br' needs to complete its configuration. */
2833 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2835 struct shash_node *port_node;
2836 struct port *port, *next;
2838 /* Get rid of deleted ports.
2839 * Get rid of deleted interfaces on ports that still exist. */
2840 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2841 port->cfg = shash_find_data(wanted_ports, port->name);
2845 port_del_ifaces(port);
2849 /* Update iface->cfg and iface->type in interfaces that still exist. */
2850 SHASH_FOR_EACH (port_node, wanted_ports) {
2851 const struct ovsrec_port *port = port_node->data;
2854 for (i = 0; i < port->n_interfaces; i++) {
2855 const struct ovsrec_interface *cfg = port->interfaces[i];
2856 struct iface *iface = iface_lookup(br, cfg->name);
2857 const char *type = iface_get_type(cfg, br->cfg);
2862 } else if (!strcmp(type, "null")) {
2863 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2864 " may be removed in February 2013. Please email"
2865 " dev@openvswitch.org with concerns.",
2868 /* We will add new interfaces later. */
2874 /* Initializes 'oc' appropriately as a management service controller for
2877 * The caller must free oc->target when it is no longer needed. */
2879 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2880 struct ofproto_controller *oc)
2882 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2883 oc->max_backoff = 0;
2884 oc->probe_interval = 60;
2885 oc->band = OFPROTO_OUT_OF_BAND;
2887 oc->burst_limit = 0;
2888 oc->enable_async_msgs = true;
2892 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2894 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2895 struct ofproto_controller *oc)
2899 oc->target = c->target;
2900 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2901 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2902 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2903 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2904 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2905 oc->burst_limit = (c->controller_burst_limit
2906 ? *c->controller_burst_limit : 0);
2907 oc->enable_async_msgs = (!c->enable_async_messages
2908 || *c->enable_async_messages);
2909 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2910 if (dscp < 0 || dscp > 63) {
2911 dscp = DSCP_DEFAULT;
2916 /* Configures the IP stack for 'br''s local interface properly according to the
2917 * configuration in 'c'. */
2919 bridge_configure_local_iface_netdev(struct bridge *br,
2920 struct ovsrec_controller *c)
2922 struct netdev *netdev;
2923 struct in_addr mask, gateway;
2925 struct iface *local_iface;
2928 /* If there's no local interface or no IP address, give up. */
2929 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2930 if (!local_iface || !c->local_ip
2931 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2935 /* Bring up the local interface. */
2936 netdev = local_iface->netdev;
2937 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2939 /* Configure the IP address and netmask. */
2940 if (!c->local_netmask
2941 || !inet_pton(AF_INET, c->local_netmask, &mask)
2943 mask.s_addr = guess_netmask(ip.s_addr);
2945 if (!netdev_set_in4(netdev, ip, mask)) {
2946 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2947 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2950 /* Configure the default gateway. */
2951 if (c->local_gateway
2952 && inet_pton(AF_INET, c->local_gateway, &gateway)
2953 && gateway.s_addr) {
2954 if (!netdev_add_router(netdev, gateway)) {
2955 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2956 br->name, IP_ARGS(gateway.s_addr));
2961 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2962 * in either string are treated as equal to any number of slashes in the other,
2963 * e.g. "x///y" is equal to "x/y".
2965 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2966 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2967 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2968 * 'b' against a prefix of 'a'.
2971 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2973 const char *b_start = b;
2975 if (b - b_start >= b_stoplen) {
2977 } else if (*a != *b) {
2979 } else if (*a == '/') {
2980 a += strspn(a, "/");
2981 b += strspn(b, "/");
2982 } else if (*a == '\0') {
2992 bridge_configure_remotes(struct bridge *br,
2993 const struct sockaddr_in *managers, size_t n_managers)
2995 bool disable_in_band;
2997 struct ovsrec_controller **controllers;
2998 size_t n_controllers;
3000 enum ofproto_fail_mode fail_mode;
3002 struct ofproto_controller *ocs;
3006 /* Check if we should disable in-band control on this bridge. */
3007 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3010 /* Set OpenFlow queue ID for in-band control. */
3011 ofproto_set_in_band_queue(br->ofproto,
3012 smap_get_int(&br->cfg->other_config,
3013 "in-band-queue", -1));
3015 if (disable_in_band) {
3016 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3018 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3021 n_controllers = bridge_get_controllers(br, &controllers);
3023 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3026 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3027 for (i = 0; i < n_controllers; i++) {
3028 struct ovsrec_controller *c = controllers[i];
3030 if (!strncmp(c->target, "punix:", 6)
3031 || !strncmp(c->target, "unix:", 5)) {
3032 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3035 if (!strncmp(c->target, "unix:", 5)) {
3036 /* Connect to a listening socket */
3037 whitelist = xasprintf("unix:%s/", ovs_rundir());
3038 if (strchr(c->target, '/') &&
3039 !equal_pathnames(c->target, whitelist,
3040 strlen(whitelist))) {
3041 /* Absolute path specified, but not in ovs_rundir */
3042 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3043 "controller \"%s\" due to possibility for "
3044 "remote exploit. Instead, specify socket "
3045 "in whitelisted \"%s\" or connect to "
3046 "\"unix:%s/%s.mgmt\" (which is always "
3047 "available without special configuration).",
3048 br->name, c->target, whitelist,
3049 ovs_rundir(), br->name);
3054 whitelist = xasprintf("punix:%s/%s.controller",
3055 ovs_rundir(), br->name);
3056 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3057 /* Prevent remote ovsdb-server users from accessing
3058 * arbitrary Unix domain sockets and overwriting arbitrary
3060 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3061 "controller \"%s\" due to possibility of "
3062 "overwriting local files. Instead, specify "
3063 "whitelisted \"%s\" or connect to "
3064 "\"unix:%s/%s.mgmt\" (which is always "
3065 "available without special configuration).",
3066 br->name, c->target, whitelist,
3067 ovs_rundir(), br->name);
3076 bridge_configure_local_iface_netdev(br, c);
3077 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3078 if (disable_in_band) {
3079 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3084 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3085 bridge_get_allowed_versions(br));
3086 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3089 /* Set the fail-mode. */
3090 fail_mode = !br->cfg->fail_mode
3091 || !strcmp(br->cfg->fail_mode, "standalone")
3092 ? OFPROTO_FAIL_STANDALONE
3093 : OFPROTO_FAIL_SECURE;
3094 ofproto_set_fail_mode(br->ofproto, fail_mode);
3096 /* Configure OpenFlow controller connection snooping. */
3097 if (!ofproto_has_snoops(br->ofproto)) {
3101 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3102 ovs_rundir(), br->name));
3103 ofproto_set_snoops(br->ofproto, &snoops);
3104 sset_destroy(&snoops);
3109 bridge_configure_tables(struct bridge *br)
3111 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3115 n_tables = ofproto_get_n_tables(br->ofproto);
3117 for (i = 0; i < n_tables; i++) {
3118 struct ofproto_table_settings s;
3121 s.max_flows = UINT_MAX;
3124 s.n_prefix_fields = 0;
3125 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3127 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3128 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3131 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3132 s.max_flows = *cfg->flow_limit;
3134 if (cfg->overflow_policy
3135 && !strcmp(cfg->overflow_policy, "evict")) {
3137 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3138 for (k = 0; k < cfg->n_groups; k++) {
3139 const char *string = cfg->groups[k];
3142 msg = mf_parse_subfield__(&s.groups[k], &string);
3144 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3145 "'groups' (%s)", br->name, i, msg);
3147 } else if (*string) {
3148 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3149 "element '%s' contains trailing garbage",
3150 br->name, i, cfg->groups[k]);
3156 /* Prefix lookup fields. */
3157 s.n_prefix_fields = 0;
3158 for (k = 0; k < cfg->n_prefixes; k++) {
3159 const char *name = cfg->prefixes[k];
3160 const struct mf_field *mf = mf_from_name(name);
3162 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3166 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3167 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3168 "%s", br->name, name);
3171 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3172 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3173 "field not used: %s", br->name, name);
3176 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3178 if (s.n_prefix_fields > 0) {
3180 struct ds ds = DS_EMPTY_INITIALIZER;
3181 for (k = 0; k < s.n_prefix_fields; k++) {
3183 ds_put_char(&ds, ',');
3185 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3187 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3188 br->name, i, ds_cstr(&ds));
3193 ofproto_configure_table(br->ofproto, i, &s);
3197 for (; j < br->cfg->n_flow_tables; j++) {
3198 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3199 "%"PRId64" not supported by this datapath", br->name,
3200 br->cfg->key_flow_tables[j]);
3205 bridge_configure_dp_desc(struct bridge *br)
3207 ofproto_set_dp_desc(br->ofproto,
3208 smap_get(&br->cfg->other_config, "dp-desc"));
3211 /* Port functions. */
3213 static struct port *
3214 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3218 port = xzalloc(sizeof *port);
3220 port->name = xstrdup(cfg->name);
3222 list_init(&port->ifaces);
3224 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3228 /* Deletes interfaces from 'port' that are no longer configured for it. */
3230 port_del_ifaces(struct port *port)
3232 struct iface *iface, *next;
3233 struct sset new_ifaces;
3236 /* Collect list of new interfaces. */
3237 sset_init(&new_ifaces);
3238 for (i = 0; i < port->cfg->n_interfaces; i++) {
3239 const char *name = port->cfg->interfaces[i]->name;
3240 const char *type = port->cfg->interfaces[i]->type;
3241 if (strcmp(type, "null")) {
3242 sset_add(&new_ifaces, name);
3246 /* Get rid of deleted interfaces. */
3247 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3248 if (!sset_contains(&new_ifaces, iface->name)) {
3249 iface_destroy(iface);
3253 sset_destroy(&new_ifaces);
3257 port_destroy(struct port *port)
3260 struct bridge *br = port->bridge;
3261 struct iface *iface, *next;
3264 ofproto_bundle_unregister(br->ofproto, port);
3267 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3268 iface_destroy__(iface);
3271 hmap_remove(&br->ports, &port->hmap_node);
3277 static struct port *
3278 port_lookup(const struct bridge *br, const char *name)
3282 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3284 if (!strcmp(port->name, name)) {
3292 enable_lacp(struct port *port, bool *activep)
3294 if (!port->cfg->lacp) {
3295 /* XXX when LACP implementation has been sufficiently tested, enable by
3296 * default and make active on bonded ports. */
3298 } else if (!strcmp(port->cfg->lacp, "off")) {
3300 } else if (!strcmp(port->cfg->lacp, "active")) {
3303 } else if (!strcmp(port->cfg->lacp, "passive")) {
3307 VLOG_WARN("port %s: unknown LACP mode %s",
3308 port->name, port->cfg->lacp);
3313 static struct lacp_settings *
3314 port_configure_lacp(struct port *port, struct lacp_settings *s)
3316 const char *lacp_time, *system_id;
3319 if (!enable_lacp(port, &s->active)) {
3323 s->name = port->name;
3325 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3327 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3328 ETH_ADDR_SCAN_ARGS(s->id))) {
3329 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3330 " address.", port->name, system_id);
3334 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3337 if (eth_addr_is_zero(s->id)) {
3338 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3342 /* Prefer bondable links if unspecified. */
3343 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3345 s->priority = (priority > 0 && priority <= UINT16_MAX
3347 : UINT16_MAX - !list_is_short(&port->ifaces));
3349 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3350 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3352 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3353 "lacp-fallback-ab", false);
3359 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3361 int priority, portid, key;
3363 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3364 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3366 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3368 if (portid <= 0 || portid > UINT16_MAX) {
3369 portid = ofp_to_u16(iface->ofp_port);
3372 if (priority <= 0 || priority > UINT16_MAX) {
3373 priority = UINT16_MAX;
3376 if (key < 0 || key > UINT16_MAX) {
3380 s->name = iface->name;
3382 s->priority = priority;
3387 port_configure_bond(struct port *port, struct bond_settings *s)
3389 const char *detect_s;
3390 struct iface *iface;
3392 int miimon_interval;
3394 s->name = port->name;
3396 if (port->cfg->bond_mode) {
3397 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3398 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3399 port->name, port->cfg->bond_mode,
3400 bond_mode_to_string(s->balance));
3403 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3405 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3406 * active-backup. At some point we should remove this warning. */
3407 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3408 " in previous versions, the default bond_mode was"
3409 " balance-slb", port->name,
3410 bond_mode_to_string(s->balance));
3412 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3413 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3414 "please use another bond type or disable flood_vlans",
3418 miimon_interval = smap_get_int(&port->cfg->other_config,
3419 "bond-miimon-interval", 0);
3420 if (miimon_interval <= 0) {
3421 miimon_interval = 200;
3424 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3425 if (!detect_s || !strcmp(detect_s, "carrier")) {
3426 miimon_interval = 0;
3427 } else if (strcmp(detect_s, "miimon")) {
3428 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3429 "defaulting to carrier", port->name, detect_s);
3430 miimon_interval = 0;
3433 s->up_delay = MAX(0, port->cfg->bond_updelay);
3434 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3435 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3436 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3437 "bond-rebalance-interval", 10000);
3438 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3439 s->rebalance_interval = 1000;
3442 s->fake_iface = port->cfg->bond_fake_iface;
3444 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3445 "lacp-fallback-ab", false);
3447 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3448 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3451 mac_s = port->cfg->bond_active_slave;
3452 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
3453 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
3454 /* OVSDB did not store the last active interface */
3455 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
3459 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3460 * instead of obtaining it from the database. */
3462 port_is_synthetic(const struct port *port)
3464 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3467 /* Interface functions. */
3470 iface_is_internal(const struct ovsrec_interface *iface,
3471 const struct ovsrec_bridge *br)
3473 /* The local port and "internal" ports are always "internal". */
3474 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3477 /* Returns the correct network device type for interface 'iface' in bridge
3480 iface_get_type(const struct ovsrec_interface *iface,
3481 const struct ovsrec_bridge *br)
3485 /* The local port always has type "internal". Other ports take
3486 * their type from the database and default to "system" if none is
3488 if (iface_is_internal(iface, br)) {
3491 type = iface->type[0] ? iface->type : "system";
3494 return ofproto_port_open_type(br->datapath_type, type);
3498 iface_destroy__(struct iface *iface)
3501 struct port *port = iface->port;
3502 struct bridge *br = port->bridge;
3504 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3505 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3508 if (iface->ofp_port != OFPP_NONE) {
3509 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3512 list_remove(&iface->port_elem);
3513 hmap_remove(&br->iface_by_name, &iface->name_node);
3515 netdev_close(iface->netdev);
3523 iface_destroy(struct iface *iface)
3526 struct port *port = iface->port;
3528 iface_destroy__(iface);
3529 if (list_is_empty(&port->ifaces)) {
3535 static struct iface *
3536 iface_lookup(const struct bridge *br, const char *name)
3538 struct iface *iface;
3540 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3541 &br->iface_by_name) {
3542 if (!strcmp(iface->name, name)) {
3550 static struct iface *
3551 iface_find(const char *name)
3553 const struct bridge *br;
3555 HMAP_FOR_EACH (br, node, &all_bridges) {
3556 struct iface *iface = iface_lookup(br, name);
3565 static struct iface *
3566 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3568 struct iface *iface;
3570 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3572 if (iface->ofp_port == ofp_port) {
3579 /* Set Ethernet address of 'iface', if one is specified in the configuration
3582 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3584 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3585 struct iface *hw_addr_iface;
3587 if (strcmp(iface->type, "internal")) {
3591 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3593 } else if (port->cfg->fake_bridge) {
3594 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3595 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3600 if (iface->ofp_port == OFPP_LOCAL) {
3601 VLOG_ERR("interface %s: ignoring mac in Interface record "
3602 "(use Bridge record to set local port's mac)",
3604 } else if (eth_addr_is_multicast(mac)) {
3605 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3608 int error = netdev_set_etheraddr(iface->netdev, mac);
3610 VLOG_ERR("interface %s: setting MAC failed (%s)",
3611 iface->name, ovs_strerror(error));
3617 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3619 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3621 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3622 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3623 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3627 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3628 * sets the "ofport" field to -1.
3630 * This is appropriate when 'if_cfg''s interface cannot be created or is
3631 * otherwise invalid. */
3633 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3635 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3636 iface_set_ofport(if_cfg, OFPP_NONE);
3637 ovsrec_interface_set_status(if_cfg, NULL);
3638 ovsrec_interface_set_admin_state(if_cfg, NULL);
3639 ovsrec_interface_set_duplex(if_cfg, NULL);
3640 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3641 ovsrec_interface_set_link_state(if_cfg, NULL);
3642 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3643 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3644 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3645 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3646 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3647 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3648 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3649 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3654 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3656 union ovsdb_atom atom;
3658 atom.integer = target;
3659 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3663 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3665 struct ofpbuf queues_buf;
3667 ofpbuf_init(&queues_buf, 0);
3669 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3670 netdev_set_qos(iface->netdev, NULL, NULL);
3672 const struct ovsdb_datum *queues;
3673 struct netdev_queue_dump dump;
3674 unsigned int queue_id;
3675 struct smap details;
3679 /* Configure top-level Qos for 'iface'. */
3680 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3682 /* Deconfigure queues that were deleted. */
3683 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3685 smap_init(&details);
3686 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3687 if (!queue_ids_include(queues, queue_id)) {
3688 netdev_delete_queue(iface->netdev, queue_id);
3691 smap_destroy(&details);
3693 /* Configure queues for 'iface'. */
3695 for (i = 0; i < qos->n_queues; i++) {
3696 const struct ovsrec_queue *queue = qos->value_queues[i];
3697 unsigned int queue_id = qos->key_queues[i];
3699 if (queue_id == 0) {
3703 if (queue->n_dscp == 1) {
3704 struct ofproto_port_queue *port_queue;
3706 port_queue = ofpbuf_put_uninit(&queues_buf,
3707 sizeof *port_queue);
3708 port_queue->queue = queue_id;
3709 port_queue->dscp = queue->dscp[0];
3712 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3715 struct smap details;
3717 smap_init(&details);
3718 netdev_set_queue(iface->netdev, 0, &details);
3719 smap_destroy(&details);
3723 if (iface->ofp_port != OFPP_NONE) {
3724 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3725 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3727 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3728 port_queues, n_queues);
3731 netdev_set_policing(iface->netdev,
3732 iface->cfg->ingress_policing_rate,
3733 iface->cfg->ingress_policing_burst);
3735 ofpbuf_uninit(&queues_buf);
3739 iface_configure_cfm(struct iface *iface)
3741 const struct ovsrec_interface *cfg = iface->cfg;
3742 const char *opstate_str;
3743 const char *cfm_ccm_vlan;
3744 struct cfm_settings s;
3745 struct smap netdev_args;
3747 if (!cfg->n_cfm_mpid) {
3748 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3752 s.check_tnl_key = false;
3753 smap_init(&netdev_args);
3754 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3755 const char *key = smap_get(&netdev_args, "key");
3756 const char *in_key = smap_get(&netdev_args, "in_key");
3758 s.check_tnl_key = (key && !strcmp(key, "flow"))
3759 || (in_key && !strcmp(in_key, "flow"));
3761 smap_destroy(&netdev_args);
3763 s.mpid = *cfg->cfm_mpid;
3764 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3765 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3766 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3768 if (s.interval <= 0) {
3772 if (!cfm_ccm_vlan) {
3774 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3775 s.ccm_vlan = CFM_RANDOM_VLAN;
3777 s.ccm_vlan = atoi(cfm_ccm_vlan);
3778 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3783 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3785 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3787 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3788 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3790 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3793 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3794 * instead of obtaining it from the database. */
3796 iface_is_synthetic(const struct iface *iface)
3798 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3802 iface_validate_ofport__(size_t n, int64_t *ofport)
3804 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3805 ? u16_to_ofp(*ofport)
3810 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3812 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3816 iface_pick_ofport(const struct ovsrec_interface *cfg)
3818 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3819 return (requested_ofport != OFPP_NONE
3821 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3824 /* Port mirroring. */
3826 static struct mirror *
3827 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3831 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3832 if (uuid_equals(uuid, &m->uuid)) {
3840 bridge_configure_mirrors(struct bridge *br)
3842 const struct ovsdb_datum *mc;
3843 unsigned long *flood_vlans;
3844 struct mirror *m, *next;
3847 /* Get rid of deleted mirrors. */
3848 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3849 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3850 union ovsdb_atom atom;
3852 atom.uuid = m->uuid;
3853 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3858 /* Add new mirrors and reconfigure existing ones. */
3859 for (i = 0; i < br->cfg->n_mirrors; i++) {
3860 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3861 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3863 m = mirror_create(br, cfg);
3866 if (!mirror_configure(m)) {
3871 /* Update flooded vlans (for RSPAN). */
3872 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3873 br->cfg->n_flood_vlans);
3874 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3875 bitmap_free(flood_vlans);
3878 static struct mirror *
3879 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3883 m = xzalloc(sizeof *m);
3884 m->uuid = cfg->header_.uuid;
3885 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3887 m->name = xstrdup(cfg->name);
3893 mirror_destroy(struct mirror *m)
3896 struct bridge *br = m->bridge;
3899 ofproto_mirror_unregister(br->ofproto, m);
3902 hmap_remove(&br->mirrors, &m->hmap_node);
3909 mirror_collect_ports(struct mirror *m,
3910 struct ovsrec_port **in_ports, int n_in_ports,
3911 void ***out_portsp, size_t *n_out_portsp)
3913 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3914 size_t n_out_ports = 0;
3917 for (i = 0; i < n_in_ports; i++) {
3918 const char *name = in_ports[i]->name;
3919 struct port *port = port_lookup(m->bridge, name);
3921 out_ports[n_out_ports++] = port;
3923 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3924 "port %s", m->bridge->name, m->name, name);
3927 *out_portsp = out_ports;
3928 *n_out_portsp = n_out_ports;
3932 mirror_configure(struct mirror *m)
3934 const struct ovsrec_mirror *cfg = m->cfg;
3935 struct ofproto_mirror_settings s;
3938 if (strcmp(cfg->name, m->name)) {
3940 m->name = xstrdup(cfg->name);
3944 /* Get output port or VLAN. */
3945 if (cfg->output_port) {
3946 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3947 if (!s.out_bundle) {
3948 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3949 m->bridge->name, m->name);
3952 s.out_vlan = UINT16_MAX;
3954 if (cfg->output_vlan) {
3955 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3956 "output vlan; ignoring output vlan",
3957 m->bridge->name, m->name);
3959 } else if (cfg->output_vlan) {
3960 /* The database should prevent invalid VLAN values. */
3961 s.out_bundle = NULL;
3962 s.out_vlan = *cfg->output_vlan;
3964 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3965 m->bridge->name, m->name);
3969 /* Get port selection. */
3970 if (cfg->select_all) {
3971 size_t n_ports = hmap_count(&m->bridge->ports);
3972 void **ports = xmalloc(n_ports * sizeof *ports);
3977 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3987 /* Get ports, dropping ports that don't exist.
3988 * The IDL ensures that there are no duplicates. */
3989 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3990 &s.srcs, &s.n_srcs);
3991 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3992 &s.dsts, &s.n_dsts);
3995 /* Get VLAN selection. */
3996 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3999 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4002 if (s.srcs != s.dsts) {
4011 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4013 * This is deprecated. It is only for compatibility with broken device drivers
4014 * in old versions of Linux that do not properly support VLANs when VLAN
4015 * devices are not used. When broken device drivers are no longer in
4016 * widespread use, we will delete these interfaces. */
4018 static struct ovsrec_port **recs;
4019 static size_t n_recs, allocated_recs;
4021 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4022 * splinters are reconfigured. */
4024 register_rec(struct ovsrec_port *rec)
4026 if (n_recs >= allocated_recs) {
4027 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4029 recs[n_recs++] = rec;
4032 /* Frees all of the ports registered with register_reg(). */
4034 free_registered_recs(void)
4038 for (i = 0; i < n_recs; i++) {
4039 struct ovsrec_port *port = recs[i];
4042 for (j = 0; j < port->n_interfaces; j++) {
4043 struct ovsrec_interface *iface = port->interfaces[j];
4048 smap_destroy(&port->other_config);
4049 free(port->interfaces);
4057 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4060 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4062 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4066 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4069 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4070 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4071 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4074 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4075 * use, returns NULL.
4077 * Updates 'vlan_splinters_enabled_anywhere'. */
4078 static unsigned long int *
4079 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4081 unsigned long int *splinter_vlans;
4082 struct sset splinter_ifaces;
4083 const char *real_dev_name;
4084 struct shash *real_devs;
4085 struct shash_node *node;
4089 /* Free space allocated for synthesized ports and interfaces, since we're
4090 * in the process of reconstructing all of them. */
4091 free_registered_recs();
4093 splinter_vlans = bitmap_allocate(4096);
4094 sset_init(&splinter_ifaces);
4095 vlan_splinters_enabled_anywhere = false;
4096 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4097 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4100 for (j = 0; j < br_cfg->n_ports; j++) {
4101 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4104 for (k = 0; k < port_cfg->n_interfaces; k++) {
4105 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4107 if (vlan_splinters_is_enabled(iface_cfg)) {
4108 vlan_splinters_enabled_anywhere = true;
4109 sset_add(&splinter_ifaces, iface_cfg->name);
4110 vlan_bitmap_from_array__(port_cfg->trunks,
4116 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4117 bitmap_set1(splinter_vlans, *port_cfg->tag);
4122 if (!vlan_splinters_enabled_anywhere) {
4123 free(splinter_vlans);
4124 sset_destroy(&splinter_ifaces);
4128 HMAP_FOR_EACH (br, node, &all_bridges) {
4130 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4134 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4135 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4136 * device to be created for it. */
4137 bitmap_set0(splinter_vlans, 0);
4138 bitmap_set0(splinter_vlans, 4095);
4140 /* Delete all VLAN devices that we don't need. */
4142 real_devs = vlandev_get_real_devs();
4143 SHASH_FOR_EACH (node, real_devs) {
4144 const struct vlan_real_dev *real_dev = node->data;
4145 const struct vlan_dev *vlan_dev;
4146 bool real_dev_has_splinters;
4148 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4150 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4151 if (!real_dev_has_splinters
4152 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4153 struct netdev *netdev;
4155 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4156 if (!netdev_get_in4(netdev, NULL, NULL) ||
4157 !netdev_get_in6(netdev, NULL)) {
4158 /* It has an IP address configured, so we don't own
4159 * it. Don't delete it. */
4161 vlandev_del(vlan_dev->name);
4163 netdev_close(netdev);
4170 /* Add all VLAN devices that we need. */
4171 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4174 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4175 if (!vlandev_get_name(real_dev_name, vid)) {
4176 vlandev_add(real_dev_name, vid);
4183 sset_destroy(&splinter_ifaces);
4185 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4186 free(splinter_vlans);
4189 return splinter_vlans;
4192 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4195 configure_splinter_port(struct port *port)
4197 struct ofproto *ofproto = port->bridge->ofproto;
4198 ofp_port_t realdev_ofp_port;
4199 const char *realdev_name;
4200 struct iface *vlandev, *realdev;
4202 ofproto_bundle_unregister(port->bridge->ofproto, port);
4204 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4207 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4208 realdev = iface_lookup(port->bridge, realdev_name);
4209 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4211 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4215 static struct ovsrec_port *
4216 synthesize_splinter_port(const char *real_dev_name,
4217 const char *vlan_dev_name, int vid)
4219 struct ovsrec_interface *iface;
4220 struct ovsrec_port *port;
4222 iface = xmalloc(sizeof *iface);
4223 ovsrec_interface_init(iface);
4224 iface->name = xstrdup(vlan_dev_name);
4225 iface->type = "system";
4227 port = xmalloc(sizeof *port);
4228 ovsrec_port_init(port);
4229 port->interfaces = xmemdup(&iface, sizeof iface);
4230 port->n_interfaces = 1;
4231 port->name = xstrdup(vlan_dev_name);
4232 port->vlan_mode = "splinter";
4233 port->tag = xmalloc(sizeof *port->tag);
4236 smap_add(&port->other_config, "realdev", real_dev_name);
4242 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4243 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4244 * 1-bit in the 'splinter_vlans' bitmap. */
4246 add_vlan_splinter_ports(struct bridge *br,
4247 const unsigned long int *splinter_vlans,
4248 struct shash *ports)
4252 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4253 * we're modifying 'ports'. */
4254 for (i = 0; i < br->cfg->n_ports; i++) {
4255 const char *name = br->cfg->ports[i]->name;
4256 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4259 for (j = 0; j < port_cfg->n_interfaces; j++) {
4260 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4262 if (vlan_splinters_is_enabled(iface_cfg)) {
4263 const char *real_dev_name;
4266 real_dev_name = iface_cfg->name;
4267 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4268 const char *vlan_dev_name;
4270 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4272 && !shash_find(ports, vlan_dev_name)) {
4273 shash_add(ports, vlan_dev_name,
4274 synthesize_splinter_port(
4275 real_dev_name, vlan_dev_name, vid));
4284 mirror_refresh_stats(struct mirror *m)
4286 struct ofproto *ofproto = m->bridge->ofproto;
4287 uint64_t tx_packets, tx_bytes;
4290 size_t stat_cnt = 0;
4292 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4293 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4297 if (tx_packets != UINT64_MAX) {
4298 keys[stat_cnt] = "tx_packets";
4299 values[stat_cnt] = tx_packets;
4302 if (tx_bytes != UINT64_MAX) {
4303 keys[stat_cnt] = "tx_bytes";
4304 values[stat_cnt] = tx_bytes;
4308 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);