1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge);
66 COVERAGE_DEFINE(bridge_reconfigure);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
81 /* These members are valid only within bridge_reconfigure(). */
82 const char *type; /* Usually same as cfg->type. */
83 const struct ovsrec_interface *cfg;
87 struct uuid uuid; /* UUID of this "mirror" record in database. */
88 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
89 struct bridge *bridge;
91 const struct ovsrec_mirror *cfg;
95 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
96 struct bridge *bridge;
99 const struct ovsrec_port *cfg;
101 /* An ordinary bridge port has 1 interface.
102 * A bridge port for bonding has at least 2 interfaces. */
103 struct list ifaces; /* List of "struct iface"s. */
107 struct hmap_node node; /* In 'all_bridges'. */
108 char *name; /* User-specified arbitrary name. */
109 char *type; /* Datapath type. */
110 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
111 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
112 const struct ovsrec_bridge *cfg;
114 /* OpenFlow switch processing. */
115 struct ofproto *ofproto; /* OpenFlow switch. */
118 struct hmap ports; /* "struct port"s indexed by name. */
119 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
120 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
122 /* Port mirroring. */
123 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
125 /* Used during reconfiguration. */
126 struct shash wanted_ports;
128 /* Synthetic local port if necessary. */
129 struct ovsrec_port synth_local_port;
130 struct ovsrec_interface synth_local_iface;
131 struct ovsrec_interface *synth_local_ifacep;
134 /* All bridges, indexed by name. */
135 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
137 /* OVSDB IDL used to obtain configuration. */
138 static struct ovsdb_idl *idl;
140 /* We want to complete daemonization, fully detaching from our parent process,
141 * only after we have completed our initial configuration, committed our state
142 * to the database, and received confirmation back from the database server
143 * that it applied the commit. This allows our parent process to know that,
144 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
145 * to have already been filled in. (It is only "very likely" rather than
146 * certain because there is always a slim possibility that the transaction will
147 * fail or that some other client has added new bridges, ports, etc. while
148 * ovs-vswitchd was configuring using an old configuration.)
150 * We only need to do this once for our initial configuration at startup, so
151 * 'initial_config_done' tracks whether we've already done it. While we are
152 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
153 * itself and is otherwise NULL. */
154 static bool initial_config_done;
155 static struct ovsdb_idl_txn *daemonize_txn;
157 /* Most recently processed IDL sequence number. */
158 static unsigned int idl_seqno;
160 /* Track changes to port connectivity. */
161 static uint64_t connectivity_seqno = LLONG_MIN;
163 /* Each time this timer expires, the bridge fetches interface and mirror
164 * statistics and pushes them into the database. */
165 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
166 static long long int iface_stats_timer = LLONG_MIN;
168 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
169 * expensive. This can cause bridge_reconfigure() to take a long time during
170 * which no other work can be done. To deal with this problem, we limit port
171 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
172 * call to bridge_reconfigure(). If there is more work to do after the limit
173 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
174 * This allows the rest of the code to catch up on important things like
175 * forwarding packets. */
176 #define OFP_PORT_ACTION_WINDOW 10
178 static void add_del_bridges(const struct ovsrec_open_vswitch *);
179 static void bridge_run__(void);
180 static void bridge_create(const struct ovsrec_bridge *);
181 static void bridge_destroy(struct bridge *);
182 static struct bridge *bridge_lookup(const char *name);
183 static unixctl_cb_func bridge_unixctl_dump_flows;
184 static unixctl_cb_func bridge_unixctl_reconnect;
185 static size_t bridge_get_controllers(const struct bridge *br,
186 struct ovsrec_controller ***controllersp);
187 static void bridge_collect_wanted_ports(struct bridge *,
188 const unsigned long *splinter_vlans,
189 struct shash *wanted_ports);
190 static void bridge_delete_ofprotos(void);
191 static void bridge_delete_or_reconfigure_ports(struct bridge *);
192 static void bridge_del_ports(struct bridge *,
193 const struct shash *wanted_ports);
194 static void bridge_add_ports(struct bridge *,
195 const struct shash *wanted_ports);
197 static void bridge_configure_datapath_id(struct bridge *);
198 static void bridge_configure_netflow(struct bridge *);
199 static void bridge_configure_forward_bpdu(struct bridge *);
200 static void bridge_configure_mac_table(struct bridge *);
201 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
202 static void bridge_configure_ipfix(struct bridge *);
203 static void bridge_configure_stp(struct bridge *);
204 static void bridge_configure_tables(struct bridge *);
205 static void bridge_configure_dp_desc(struct bridge *);
206 static void bridge_configure_remotes(struct bridge *,
207 const struct sockaddr_in *managers,
209 static void bridge_pick_local_hw_addr(struct bridge *,
210 uint8_t ea[ETH_ADDR_LEN],
211 struct iface **hw_addr_iface);
212 static uint64_t bridge_pick_datapath_id(struct bridge *,
213 const uint8_t bridge_ea[ETH_ADDR_LEN],
214 struct iface *hw_addr_iface);
215 static uint64_t dpid_from_hash(const void *, size_t nbytes);
216 static bool bridge_has_bond_fake_iface(const struct bridge *,
218 static bool port_is_bond_fake_iface(const struct port *);
220 static unixctl_cb_func qos_unixctl_show;
222 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
223 static void port_del_ifaces(struct port *);
224 static void port_destroy(struct port *);
225 static struct port *port_lookup(const struct bridge *, const char *name);
226 static void port_configure(struct port *);
227 static struct lacp_settings *port_configure_lacp(struct port *,
228 struct lacp_settings *);
229 static void port_configure_bond(struct port *, struct bond_settings *);
230 static bool port_is_synthetic(const struct port *);
232 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
233 static void run_system_stats(void);
235 static void bridge_configure_mirrors(struct bridge *);
236 static struct mirror *mirror_create(struct bridge *,
237 const struct ovsrec_mirror *);
238 static void mirror_destroy(struct mirror *);
239 static bool mirror_configure(struct mirror *);
240 static void mirror_refresh_stats(struct mirror *);
242 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
243 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
244 const struct ovsrec_port *);
245 static bool iface_is_internal(const struct ovsrec_interface *iface,
246 const struct ovsrec_bridge *br);
247 static const char *iface_get_type(const struct ovsrec_interface *,
248 const struct ovsrec_bridge *);
249 static void iface_destroy(struct iface *);
250 static struct iface *iface_lookup(const struct bridge *, const char *name);
251 static struct iface *iface_find(const char *name);
252 static struct iface *iface_from_ofp_port(const struct bridge *,
253 ofp_port_t ofp_port);
254 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
255 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
256 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
257 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
258 static void iface_configure_cfm(struct iface *);
259 static void iface_refresh_cfm_stats(struct iface *);
260 static void iface_refresh_stats(struct iface *);
261 static void iface_refresh_status(struct iface *);
262 static bool iface_is_synthetic(const struct iface *);
263 static ofp_port_t iface_get_requested_ofp_port(
264 const struct ovsrec_interface *);
265 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
267 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
269 * This is deprecated. It is only for compatibility with broken device drivers
270 * in old versions of Linux that do not properly support VLANs when VLAN
271 * devices are not used. When broken device drivers are no longer in
272 * widespread use, we will delete these interfaces. */
274 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
275 static bool vlan_splinters_enabled_anywhere;
277 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
278 static unsigned long int *collect_splinter_vlans(
279 const struct ovsrec_open_vswitch *);
280 static void configure_splinter_port(struct port *);
281 static void add_vlan_splinter_ports(struct bridge *,
282 const unsigned long int *splinter_vlans,
283 struct shash *ports);
286 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
288 struct shash iface_hints;
289 static bool initialized = false;
296 shash_init(&iface_hints);
299 for (i = 0; i < cfg->n_bridges; i++) {
300 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
303 for (j = 0; j < br_cfg->n_ports; j++) {
304 struct ovsrec_port *port_cfg = br_cfg->ports[j];
307 for (k = 0; k < port_cfg->n_interfaces; k++) {
308 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
309 struct iface_hint *iface_hint;
311 iface_hint = xmalloc(sizeof *iface_hint);
312 iface_hint->br_name = br_cfg->name;
313 iface_hint->br_type = br_cfg->datapath_type;
314 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
316 shash_add(&iface_hints, if_cfg->name, iface_hint);
322 ofproto_init(&iface_hints);
324 shash_destroy_free_data(&iface_hints);
328 /* Public functions. */
330 /* Initializes the bridge module, configuring it to obtain its configuration
331 * from an OVSDB server accessed over 'remote', which should be a string in a
332 * form acceptable to ovsdb_idl_create(). */
334 bridge_init(const char *remote)
336 /* Create connection to database. */
337 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
338 idl_seqno = ovsdb_idl_get_seqno(idl);
339 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
340 ovsdb_idl_verify_write_only(idl);
342 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
343 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
344 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
345 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
350 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
351 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
352 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
354 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
355 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
356 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
358 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
377 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
382 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
384 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
386 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
389 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
391 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
402 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
404 /* Register unixctl commands. */
405 unixctl_command_register("qos/show", "interface", 1, 1,
406 qos_unixctl_show, NULL);
407 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
408 bridge_unixctl_dump_flows, NULL);
409 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
410 bridge_unixctl_reconnect, NULL);
420 struct bridge *br, *next_br;
422 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
425 ovsdb_idl_destroy(idl);
428 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
429 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
430 * responsible for freeing '*managersp' (with free()).
432 * You may be asking yourself "why does ovs-vswitchd care?", because
433 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
434 * should not be and in fact is not directly involved in that. But
435 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
436 * it has to tell in-band control where the managers are to enable that.
437 * (Thus, only managers connected in-band are collected.)
440 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
441 struct sockaddr_in **managersp, size_t *n_managersp)
443 struct sockaddr_in *managers = NULL;
444 size_t n_managers = 0;
448 /* Collect all of the potential targets from the "targets" columns of the
449 * rows pointed to by "manager_options", excluding any that are
452 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
453 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
455 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
456 sset_find_and_delete(&targets, m->target);
458 sset_add(&targets, m->target);
462 /* Now extract the targets' IP addresses. */
463 if (!sset_is_empty(&targets)) {
466 managers = xmalloc(sset_count(&targets) * sizeof *managers);
467 SSET_FOR_EACH (target, &targets) {
468 struct sockaddr_storage ss;
470 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
472 && ss.ss_family == AF_INET) {
473 managers[n_managers++] = *(struct sockaddr_in *) &ss;
477 sset_destroy(&targets);
479 *managersp = managers;
480 *n_managersp = n_managers;
484 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
486 unsigned long int *splinter_vlans;
487 struct sockaddr_in *managers;
488 struct bridge *br, *next;
489 int sflow_bridge_number;
492 COVERAGE_INC(bridge_reconfigure);
494 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
495 OFPROTO_FLOW_LIMIT_DEFAULT));
496 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
497 OFPROTO_MAX_IDLE_DEFAULT));
500 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
501 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
503 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
504 * to 'ovs_cfg', with only very minimal configuration otherwise.
506 * This is mostly an update to bridge data structures. Nothing is pushed
507 * down to ofproto or lower layers. */
508 add_del_bridges(ovs_cfg);
509 splinter_vlans = collect_splinter_vlans(ovs_cfg);
510 HMAP_FOR_EACH (br, node, &all_bridges) {
511 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
512 bridge_del_ports(br, &br->wanted_ports);
514 free(splinter_vlans);
516 /* Start pushing configuration changes down to the ofproto layer:
518 * - Delete ofprotos that are no longer configured.
520 * - Delete ports that are no longer configured.
522 * - Reconfigure existing ports to their desired configurations, or
523 * delete them if not possible.
525 * We have to do all the deletions before we can do any additions, because
526 * the ports to be added might require resources that will be freed up by
527 * deletions (they might especially overlap in name). */
528 bridge_delete_ofprotos();
529 HMAP_FOR_EACH (br, node, &all_bridges) {
531 bridge_delete_or_reconfigure_ports(br);
535 /* Finish pushing configuration changes to the ofproto layer:
537 * - Create ofprotos that are missing.
539 * - Add ports that are missing. */
540 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
544 error = ofproto_create(br->name, br->type, &br->ofproto);
546 VLOG_ERR("failed to create bridge %s: %s", br->name,
547 ovs_strerror(error));
548 shash_destroy(&br->wanted_ports);
553 HMAP_FOR_EACH (br, node, &all_bridges) {
554 bridge_add_ports(br, &br->wanted_ports);
555 shash_destroy(&br->wanted_ports);
558 reconfigure_system_stats(ovs_cfg);
560 /* Complete the configuration. */
561 sflow_bridge_number = 0;
562 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
563 HMAP_FOR_EACH (br, node, &all_bridges) {
566 /* We need the datapath ID early to allow LACP ports to use it as the
567 * default system ID. */
568 bridge_configure_datapath_id(br);
570 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
573 port_configure(port);
575 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
576 iface_set_ofport(iface->cfg, iface->ofp_port);
577 iface_configure_cfm(iface);
578 iface_configure_qos(iface, port->cfg->qos);
579 iface_set_mac(br, port, iface);
580 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
584 bridge_configure_mirrors(br);
585 bridge_configure_forward_bpdu(br);
586 bridge_configure_mac_table(br);
587 bridge_configure_remotes(br, managers, n_managers);
588 bridge_configure_netflow(br);
589 bridge_configure_sflow(br, &sflow_bridge_number);
590 bridge_configure_ipfix(br);
591 bridge_configure_stp(br);
592 bridge_configure_tables(br);
593 bridge_configure_dp_desc(br);
595 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
596 /* XXX: Remove this warning message eventually. */
597 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
598 " moved to the Open_vSwitch table. Ignoring its"
599 " setting in the bridge table.");
604 /* The ofproto-dpif provider does some final reconfiguration in its
605 * ->type_run() function. We have to call it before notifying the database
606 * client that reconfiguration is complete, otherwise there is a very
607 * narrow race window in which e.g. ofproto/trace will not recognize the
608 * new configuration (sometimes this causes unit test failures). */
612 /* Delete ofprotos which aren't configured or have the wrong type. Create
613 * ofprotos which don't exist but need to. */
615 bridge_delete_ofprotos(void)
622 /* Delete ofprotos with no bridge or with the wrong type. */
625 ofproto_enumerate_types(&types);
626 SSET_FOR_EACH (type, &types) {
629 ofproto_enumerate_names(type, &names);
630 SSET_FOR_EACH (name, &names) {
631 br = bridge_lookup(name);
632 if (!br || strcmp(type, br->type)) {
633 ofproto_delete(name, type);
637 sset_destroy(&names);
638 sset_destroy(&types);
642 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
644 if (*n >= *allocated) {
645 ports = x2nrealloc(ports, allocated, sizeof *ports);
647 ports[(*n)++] = port;
652 bridge_delete_or_reconfigure_ports(struct bridge *br)
654 struct ofproto_port ofproto_port;
655 struct ofproto_port_dump dump;
657 /* List of "ofp_port"s to delete. We make a list instead of deleting them
658 * right away because ofproto implementations aren't necessarily able to
659 * iterate through a changing list of ports in an entirely robust way. */
667 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
668 ofp_port_t requested_ofp_port;
671 iface = iface_lookup(br, ofproto_port.name);
673 /* No such iface is configured, so we should delete this
676 * As a corner case exception, keep the port if it's a bond fake
678 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
679 && !strcmp(ofproto_port.type, "internal")) {
685 if (strcmp(ofproto_port.type, iface->type)
686 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
687 /* The interface is the wrong type or can't be configured.
692 /* If the requested OpenFlow port for 'iface' changed, and it's not
693 * already the correct port, then we might want to temporarily delete
694 * this interface, so we can add it back again with the new OpenFlow
696 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
697 if (iface->ofp_port != OFPP_LOCAL &&
698 requested_ofp_port != OFPP_NONE &&
699 requested_ofp_port != iface->ofp_port) {
700 ofp_port_t victim_request;
701 struct iface *victim;
703 /* Check for an existing OpenFlow port currently occupying
704 * 'iface''s requested port number. If there isn't one, then
705 * delete this port. Otherwise we need to consider further. */
706 victim = iface_from_ofp_port(br, requested_ofp_port);
711 /* 'victim' is a port currently using 'iface''s requested port
712 * number. Unless 'victim' specifically requested that port
713 * number, too, then we can delete both 'iface' and 'victim'
714 * temporarily. (We'll add both of them back again later with new
715 * OpenFlow port numbers.)
717 * If 'victim' did request port number 'requested_ofp_port', just
718 * like 'iface', then that's a configuration inconsistency that we
719 * can't resolve. We might as well let it keep its current port
721 victim_request = iface_get_requested_ofp_port(victim->cfg);
722 if (victim_request != requested_ofp_port) {
723 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
724 iface_destroy(victim);
733 iface_destroy(iface);
734 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
737 for (i = 0; i < n; i++) {
738 ofproto_port_del(br->ofproto, del[i]);
744 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
745 bool with_requested_port)
747 struct shash_node *port_node;
749 SHASH_FOR_EACH (port_node, wanted_ports) {
750 const struct ovsrec_port *port_cfg = port_node->data;
753 for (i = 0; i < port_cfg->n_interfaces; i++) {
754 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
755 ofp_port_t requested_ofp_port;
757 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
758 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
759 struct iface *iface = iface_lookup(br, iface_cfg->name);
762 iface_create(br, iface_cfg, port_cfg);
770 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
772 /* First add interfaces that request a particular port number. */
773 bridge_add_ports__(br, wanted_ports, true);
775 /* Then add interfaces that want automatic port number assignment.
776 * We add these afterward to avoid accidentally taking a specifically
777 * requested port number. */
778 bridge_add_ports__(br, wanted_ports, false);
782 port_configure(struct port *port)
784 const struct ovsrec_port *cfg = port->cfg;
785 struct bond_settings bond_settings;
786 struct lacp_settings lacp_settings;
787 struct ofproto_bundle_settings s;
790 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
791 configure_splinter_port(port);
800 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
801 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
802 s.slaves[s.n_slaves++] = iface->ofp_port;
807 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
811 /* Get VLAN trunks. */
814 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
818 if (cfg->vlan_mode) {
819 if (!strcmp(cfg->vlan_mode, "access")) {
820 s.vlan_mode = PORT_VLAN_ACCESS;
821 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
822 s.vlan_mode = PORT_VLAN_TRUNK;
823 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
824 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
825 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
826 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
828 /* This "can't happen" because ovsdb-server should prevent it. */
829 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
830 s.vlan_mode = PORT_VLAN_TRUNK;
834 s.vlan_mode = PORT_VLAN_ACCESS;
836 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
840 s.vlan_mode = PORT_VLAN_TRUNK;
843 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
846 /* Get LACP settings. */
847 s.lacp = port_configure_lacp(port, &lacp_settings);
851 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
852 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
853 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
856 s.lacp_slaves = NULL;
859 /* Get bond settings. */
860 if (s.n_slaves > 1) {
861 s.bond = &bond_settings;
862 port_configure_bond(port, &bond_settings);
865 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
866 netdev_set_miimon_interval(iface->netdev, 0);
871 ofproto_bundle_register(port->bridge->ofproto, port, &s);
879 /* Pick local port hardware address and datapath ID for 'br'. */
881 bridge_configure_datapath_id(struct bridge *br)
883 uint8_t ea[ETH_ADDR_LEN];
885 struct iface *local_iface;
886 struct iface *hw_addr_iface;
889 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
890 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
892 int error = netdev_set_etheraddr(local_iface->netdev, ea);
894 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
895 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
896 "Ethernet address: %s",
897 br->name, ovs_strerror(error));
900 memcpy(br->ea, ea, ETH_ADDR_LEN);
902 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
903 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
904 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
905 ofproto_set_datapath_id(br->ofproto, dpid);
908 dpid_string = xasprintf("%016"PRIx64, dpid);
909 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
913 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
916 bridge_get_allowed_versions(struct bridge *br)
918 if (!br->cfg->n_protocols)
921 return ofputil_versions_from_strings(br->cfg->protocols,
922 br->cfg->n_protocols);
925 /* Set NetFlow configuration on 'br'. */
927 bridge_configure_netflow(struct bridge *br)
929 struct ovsrec_netflow *cfg = br->cfg->netflow;
930 struct netflow_options opts;
933 ofproto_set_netflow(br->ofproto, NULL);
937 memset(&opts, 0, sizeof opts);
939 /* Get default NetFlow configuration from datapath.
940 * Apply overrides from 'cfg'. */
941 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
942 if (cfg->engine_type) {
943 opts.engine_type = *cfg->engine_type;
945 if (cfg->engine_id) {
946 opts.engine_id = *cfg->engine_id;
949 /* Configure active timeout interval. */
950 opts.active_timeout = cfg->active_timeout;
951 if (!opts.active_timeout) {
952 opts.active_timeout = -1;
953 } else if (opts.active_timeout < 0) {
954 VLOG_WARN("bridge %s: active timeout interval set to negative "
955 "value, using default instead (%d seconds)", br->name,
956 NF_ACTIVE_TIMEOUT_DEFAULT);
957 opts.active_timeout = -1;
960 /* Add engine ID to interface number to disambiguate bridgs? */
961 opts.add_id_to_iface = cfg->add_id_to_interface;
962 if (opts.add_id_to_iface) {
963 if (opts.engine_id > 0x7f) {
964 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
965 "another vswitch, choose an engine id less than 128",
968 if (hmap_count(&br->ports) > 508) {
969 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
970 "another port when more than 508 ports are used",
976 sset_init(&opts.collectors);
977 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
980 if (ofproto_set_netflow(br->ofproto, &opts)) {
981 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
983 sset_destroy(&opts.collectors);
986 /* Set sFlow configuration on 'br'. */
988 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
990 const struct ovsrec_sflow *cfg = br->cfg->sflow;
991 struct ovsrec_controller **controllers;
992 struct ofproto_sflow_options oso;
993 size_t n_controllers;
997 ofproto_set_sflow(br->ofproto, NULL);
1001 memset(&oso, 0, sizeof oso);
1003 sset_init(&oso.targets);
1004 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1006 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1007 if (cfg->sampling) {
1008 oso.sampling_rate = *cfg->sampling;
1011 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1013 oso.polling_interval = *cfg->polling;
1016 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1018 oso.header_len = *cfg->header;
1021 oso.sub_id = (*sflow_bridge_number)++;
1022 oso.agent_device = cfg->agent;
1024 oso.control_ip = NULL;
1025 n_controllers = bridge_get_controllers(br, &controllers);
1026 for (i = 0; i < n_controllers; i++) {
1027 if (controllers[i]->local_ip) {
1028 oso.control_ip = controllers[i]->local_ip;
1032 ofproto_set_sflow(br->ofproto, &oso);
1034 sset_destroy(&oso.targets);
1037 /* Returns whether a IPFIX row is valid. */
1039 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1041 return ipfix && ipfix->n_targets > 0;
1044 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1046 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1047 const struct bridge *br)
1049 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1052 /* Set IPFIX configuration on 'br'. */
1054 bridge_configure_ipfix(struct bridge *br)
1056 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1057 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1058 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1059 struct ofproto_ipfix_bridge_exporter_options be_opts;
1060 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1061 size_t n_fe_opts = 0;
1063 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1064 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1069 if (!valid_be_cfg && n_fe_opts == 0) {
1070 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1075 memset(&be_opts, 0, sizeof be_opts);
1077 sset_init(&be_opts.targets);
1078 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1080 if (be_cfg->sampling) {
1081 be_opts.sampling_rate = *be_cfg->sampling;
1083 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1085 if (be_cfg->obs_domain_id) {
1086 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1088 if (be_cfg->obs_point_id) {
1089 be_opts.obs_point_id = *be_cfg->obs_point_id;
1091 if (be_cfg->cache_active_timeout) {
1092 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1094 if (be_cfg->cache_max_flows) {
1095 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1099 if (n_fe_opts > 0) {
1100 struct ofproto_ipfix_flow_exporter_options *opts;
1101 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1103 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1104 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1105 opts->collector_set_id = fe_cfg->id;
1106 sset_init(&opts->targets);
1107 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1108 fe_cfg->ipfix->n_targets);
1109 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1110 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1111 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1112 ? *fe_cfg->ipfix->cache_max_flows : 0;
1118 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1122 sset_destroy(&be_opts.targets);
1125 if (n_fe_opts > 0) {
1126 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1128 for (i = 0; i < n_fe_opts; i++) {
1129 sset_destroy(&opts->targets);
1137 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1138 struct ofproto_port_stp_settings *port_s,
1139 int *port_num_counter, unsigned long *port_num_bitmap)
1141 const char *config_str;
1142 struct iface *iface;
1144 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1145 port_s->enable = false;
1148 port_s->enable = true;
1151 /* STP over bonds is not supported. */
1152 if (!list_is_singleton(&port->ifaces)) {
1153 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1155 port_s->enable = false;
1159 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1161 /* Internal ports shouldn't participate in spanning tree, so
1163 if (!strcmp(iface->type, "internal")) {
1164 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1165 port_s->enable = false;
1169 /* STP on mirror output ports is not supported. */
1170 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1171 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1172 port_s->enable = false;
1176 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1178 unsigned long int port_num = strtoul(config_str, NULL, 0);
1179 int port_idx = port_num - 1;
1181 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1182 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1183 port_s->enable = false;
1187 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1188 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1189 port->name, port_num);
1190 port_s->enable = false;
1193 bitmap_set1(port_num_bitmap, port_idx);
1194 port_s->port_num = port_idx;
1196 if (*port_num_counter >= STP_MAX_PORTS) {
1197 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1198 port_s->enable = false;
1202 port_s->port_num = (*port_num_counter)++;
1205 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1207 port_s->path_cost = strtoul(config_str, NULL, 10);
1209 enum netdev_features current;
1212 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1213 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1214 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1217 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1219 port_s->priority = strtoul(config_str, NULL, 0);
1221 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1225 /* Set spanning tree configuration on 'br'. */
1227 bridge_configure_stp(struct bridge *br)
1229 if (!br->cfg->stp_enable) {
1230 ofproto_set_stp(br->ofproto, NULL);
1232 struct ofproto_stp_settings br_s;
1233 const char *config_str;
1235 int port_num_counter;
1236 unsigned long *port_num_bitmap;
1238 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1240 uint8_t ea[ETH_ADDR_LEN];
1242 if (eth_addr_from_string(config_str, ea)) {
1243 br_s.system_id = eth_addr_to_uint64(ea);
1245 br_s.system_id = eth_addr_to_uint64(br->ea);
1246 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1247 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1250 br_s.system_id = eth_addr_to_uint64(br->ea);
1253 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1255 br_s.priority = strtoul(config_str, NULL, 0);
1257 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1260 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1262 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1264 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1267 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1269 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1271 br_s.max_age = STP_DEFAULT_MAX_AGE;
1274 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1276 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1278 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1281 /* Configure STP on the bridge. */
1282 if (ofproto_set_stp(br->ofproto, &br_s)) {
1283 VLOG_ERR("bridge %s: could not enable STP", br->name);
1287 /* Users must either set the port number with the "stp-port-num"
1288 * configuration on all ports or none. If manual configuration
1289 * is not done, then we allocate them sequentially. */
1290 port_num_counter = 0;
1291 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1292 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1293 struct ofproto_port_stp_settings port_s;
1294 struct iface *iface;
1296 port_configure_stp(br->ofproto, port, &port_s,
1297 &port_num_counter, port_num_bitmap);
1299 /* As bonds are not supported, just apply configuration to
1300 * all interfaces. */
1301 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1302 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1304 VLOG_ERR("port %s: could not enable STP", port->name);
1310 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1311 && port_num_counter) {
1312 VLOG_ERR("bridge %s: must manually configure all STP port "
1313 "IDs or none, disabling", br->name);
1314 ofproto_set_stp(br->ofproto, NULL);
1316 bitmap_free(port_num_bitmap);
1321 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1323 const struct port *port = port_lookup(br, name);
1324 return port && port_is_bond_fake_iface(port);
1328 port_is_bond_fake_iface(const struct port *port)
1330 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1334 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1336 struct bridge *br, *next;
1337 struct shash new_br;
1340 /* Collect new bridges' names and types. */
1341 shash_init(&new_br);
1342 for (i = 0; i < cfg->n_bridges; i++) {
1343 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1344 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1346 if (strchr(br_cfg->name, '/')) {
1347 /* Prevent remote ovsdb-server users from accessing arbitrary
1348 * directories, e.g. consider a bridge named "../../../etc/". */
1349 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1351 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1352 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1356 /* Get rid of deleted bridges or those whose types have changed.
1357 * Update 'cfg' of bridges that still exist. */
1358 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1359 br->cfg = shash_find_data(&new_br, br->name);
1360 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1361 br->cfg->datapath_type))) {
1366 /* Add new bridges. */
1367 for (i = 0; i < cfg->n_bridges; i++) {
1368 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1369 struct bridge *br = bridge_lookup(br_cfg->name);
1371 bridge_create(br_cfg);
1375 shash_destroy(&new_br);
1378 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1379 * Returns 0 if successful, otherwise a positive errno value. */
1381 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1382 struct netdev *netdev)
1384 return netdev_set_config(netdev, &iface_cfg->options);
1387 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1388 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1390 * If successful, returns 0 and stores the network device in '*netdevp'. On
1391 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1393 iface_do_create(const struct bridge *br,
1394 const struct ovsrec_interface *iface_cfg,
1395 const struct ovsrec_port *port_cfg,
1396 ofp_port_t *ofp_portp, struct netdev **netdevp)
1398 struct netdev *netdev = NULL;
1401 if (netdev_is_reserved_name(iface_cfg->name)) {
1402 VLOG_WARN("could not create interface %s, name is reserved",
1408 error = netdev_open(iface_cfg->name,
1409 iface_get_type(iface_cfg, br->cfg), &netdev);
1411 VLOG_WARN("could not open network device %s (%s)",
1412 iface_cfg->name, ovs_strerror(error));
1416 error = iface_set_netdev_config(iface_cfg, netdev);
1421 *ofp_portp = iface_pick_ofport(iface_cfg);
1422 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1427 VLOG_INFO("bridge %s: added interface %s on port %d",
1428 br->name, iface_cfg->name, *ofp_portp);
1430 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1431 || iface_is_internal(iface_cfg, br->cfg)) {
1432 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1440 netdev_close(netdev);
1444 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1445 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1446 * automatically allocated for the iface. Takes ownership of and
1447 * deallocates 'if_cfg'.
1449 * Return true if an iface is successfully created, false otherwise. */
1451 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1452 const struct ovsrec_port *port_cfg)
1454 struct netdev *netdev;
1455 struct iface *iface;
1456 ofp_port_t ofp_port;
1460 /* Do the bits that can fail up front. */
1461 ovs_assert(!iface_lookup(br, iface_cfg->name));
1462 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1464 iface_clear_db_record(iface_cfg);
1468 /* Get or create the port structure. */
1469 port = port_lookup(br, port_cfg->name);
1471 port = port_create(br, port_cfg);
1474 /* Create the iface structure. */
1475 iface = xzalloc(sizeof *iface);
1476 list_push_back(&port->ifaces, &iface->port_elem);
1477 hmap_insert(&br->iface_by_name, &iface->name_node,
1478 hash_string(iface_cfg->name, 0));
1480 iface->name = xstrdup(iface_cfg->name);
1481 iface->ofp_port = ofp_port;
1482 iface->netdev = netdev;
1483 iface->type = iface_get_type(iface_cfg, br->cfg);
1484 iface->cfg = iface_cfg;
1485 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1486 hash_ofp_port(ofp_port));
1488 /* Populate initial status in database. */
1489 iface_refresh_stats(iface);
1490 iface_refresh_status(iface);
1492 /* Add bond fake iface if necessary. */
1493 if (port_is_bond_fake_iface(port)) {
1494 struct ofproto_port ofproto_port;
1496 if (ofproto_port_query_by_name(br->ofproto, port->name,
1498 struct netdev *netdev;
1501 error = netdev_open(port->name, "internal", &netdev);
1503 ofp_port_t fake_ofp_port = OFPP_NONE;
1504 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1505 netdev_close(netdev);
1507 VLOG_WARN("could not open network device %s (%s)",
1508 port->name, ovs_strerror(error));
1511 /* Already exists, nothing to do. */
1512 ofproto_port_destroy(&ofproto_port);
1519 /* Set forward BPDU option. */
1521 bridge_configure_forward_bpdu(struct bridge *br)
1523 ofproto_set_forward_bpdu(br->ofproto,
1524 smap_get_bool(&br->cfg->other_config,
1529 /* Set MAC learning table configuration for 'br'. */
1531 bridge_configure_mac_table(struct bridge *br)
1533 const char *idle_time_str;
1536 const char *mac_table_size_str;
1539 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1540 idle_time = (idle_time_str && atoi(idle_time_str)
1541 ? atoi(idle_time_str)
1542 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1544 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1545 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1546 ? atoi(mac_table_size_str)
1549 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1553 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1554 const struct port *fake_br, struct iface **hw_addr_iface)
1556 struct hmapx mirror_output_ports;
1558 bool found_addr = false;
1562 /* Mirror output ports don't participate in picking the local hardware
1563 * address. ofproto can't help us find out whether a given port is a
1564 * mirror output because we haven't configured mirrors yet, so we need to
1565 * accumulate them ourselves. */
1566 hmapx_init(&mirror_output_ports);
1567 for (i = 0; i < br->cfg->n_mirrors; i++) {
1568 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1569 if (m->output_port) {
1570 hmapx_add(&mirror_output_ports, m->output_port);
1574 /* Otherwise choose the minimum non-local MAC address among all of the
1576 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1577 uint8_t iface_ea[ETH_ADDR_LEN];
1578 struct iface *candidate;
1579 struct iface *iface;
1581 /* Mirror output ports don't participate. */
1582 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1586 /* Choose the MAC address to represent the port. */
1588 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1589 /* Find the interface with this Ethernet address (if any) so that
1590 * we can provide the correct devname to the caller. */
1591 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1592 uint8_t candidate_ea[ETH_ADDR_LEN];
1593 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1594 && eth_addr_equals(iface_ea, candidate_ea)) {
1599 /* Choose the interface whose MAC address will represent the port.
1600 * The Linux kernel bonding code always chooses the MAC address of
1601 * the first slave added to a bond, and the Fedora networking
1602 * scripts always add slaves to a bond in alphabetical order, so
1603 * for compatibility we choose the interface with the name that is
1604 * first in alphabetical order. */
1605 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1606 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1611 /* The local port doesn't count (since we're trying to choose its
1612 * MAC address anyway). */
1613 if (iface->ofp_port == OFPP_LOCAL) {
1617 /* For fake bridges we only choose from ports with the same tag */
1618 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1619 if (!port->cfg->tag) {
1622 if (*port->cfg->tag != *fake_br->cfg->tag) {
1628 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1634 /* Compare against our current choice. */
1635 if (!eth_addr_is_multicast(iface_ea) &&
1636 !eth_addr_is_local(iface_ea) &&
1637 !eth_addr_is_reserved(iface_ea) &&
1638 !eth_addr_is_zero(iface_ea) &&
1639 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1641 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1642 *hw_addr_iface = iface;
1648 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1649 *hw_addr_iface = NULL;
1652 hmapx_destroy(&mirror_output_ports);
1656 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1657 struct iface **hw_addr_iface)
1660 *hw_addr_iface = NULL;
1662 /* Did the user request a particular MAC? */
1663 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1664 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1665 if (eth_addr_is_multicast(ea)) {
1666 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1667 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1668 } else if (eth_addr_is_zero(ea)) {
1669 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1675 /* Find a local hw address */
1676 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1679 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1680 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1681 * an interface on 'br', then that interface must be passed in as
1682 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1683 * 'hw_addr_iface' must be passed in as a null pointer. */
1685 bridge_pick_datapath_id(struct bridge *br,
1686 const uint8_t bridge_ea[ETH_ADDR_LEN],
1687 struct iface *hw_addr_iface)
1690 * The procedure for choosing a bridge MAC address will, in the most
1691 * ordinary case, also choose a unique MAC that we can use as a datapath
1692 * ID. In some special cases, though, multiple bridges will end up with
1693 * the same MAC address. This is OK for the bridges, but it will confuse
1694 * the OpenFlow controller, because each datapath needs a unique datapath
1697 * Datapath IDs must be unique. It is also very desirable that they be
1698 * stable from one run to the next, so that policy set on a datapath
1701 const char *datapath_id;
1704 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1705 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1709 if (!hw_addr_iface) {
1711 * A purely internal bridge, that is, one that has no non-virtual
1712 * network devices on it at all, is difficult because it has no
1713 * natural unique identifier at all.
1715 * When the host is a XenServer, we handle this case by hashing the
1716 * host's UUID with the name of the bridge. Names of bridges are
1717 * persistent across XenServer reboots, although they can be reused if
1718 * an internal network is destroyed and then a new one is later
1719 * created, so this is fairly effective.
1721 * When the host is not a XenServer, we punt by using a random MAC
1722 * address on each run.
1724 const char *host_uuid = xenserver_get_host_uuid();
1726 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1727 dpid = dpid_from_hash(combined, strlen(combined));
1733 return eth_addr_to_uint64(bridge_ea);
1737 dpid_from_hash(const void *data, size_t n)
1739 uint8_t hash[SHA1_DIGEST_SIZE];
1741 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1742 sha1_bytes(data, n, hash);
1743 eth_addr_mark_random(hash);
1744 return eth_addr_to_uint64(hash);
1748 iface_refresh_status(struct iface *iface)
1752 enum netdev_features current;
1756 uint8_t mac[ETH_ADDR_LEN];
1760 if (iface_is_synthetic(iface)) {
1766 if (!netdev_get_status(iface->netdev, &smap)) {
1767 ovsrec_interface_set_status(iface->cfg, &smap);
1769 ovsrec_interface_set_status(iface->cfg, NULL);
1772 smap_destroy(&smap);
1774 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1775 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1777 ovsrec_interface_set_duplex(iface->cfg,
1778 netdev_features_is_full_duplex(current)
1780 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1782 ovsrec_interface_set_duplex(iface->cfg, NULL);
1783 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1786 error = netdev_get_mtu(iface->netdev, &mtu);
1789 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1791 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1794 error = netdev_get_etheraddr(iface->netdev, mac);
1796 char mac_string[32];
1798 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1799 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1801 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1804 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1805 * if there is no valid ifindex number. */
1806 ifindex64 = netdev_get_ifindex(iface->netdev);
1807 if (ifindex64 < 0) {
1810 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1813 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1816 iface_refresh_cfm_stats(struct iface *iface)
1818 const struct ovsrec_interface *cfg = iface->cfg;
1819 struct ofproto_cfm_status status;
1821 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1822 iface->ofp_port, &status)) {
1823 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1824 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1825 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1826 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1827 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1828 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1830 const char *reasons[CFM_FAULT_N_REASONS];
1831 int64_t cfm_health = status.health;
1832 int64_t cfm_flap_count = status.flap_count;
1833 bool faulted = status.faults != 0;
1836 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1839 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1840 int reason = 1 << i;
1841 if (status.faults & reason) {
1842 reasons[j++] = cfm_fault_reason_to_str(reason);
1845 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1847 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1849 if (status.remote_opstate >= 0) {
1850 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1851 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1853 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1856 ovsrec_interface_set_cfm_remote_mpids(cfg,
1857 (const int64_t *)status.rmps,
1859 if (cfm_health >= 0) {
1860 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1862 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1870 iface_refresh_stats(struct iface *iface)
1872 #define IFACE_STATS \
1873 IFACE_STAT(rx_packets, "rx_packets") \
1874 IFACE_STAT(tx_packets, "tx_packets") \
1875 IFACE_STAT(rx_bytes, "rx_bytes") \
1876 IFACE_STAT(tx_bytes, "tx_bytes") \
1877 IFACE_STAT(rx_dropped, "rx_dropped") \
1878 IFACE_STAT(tx_dropped, "tx_dropped") \
1879 IFACE_STAT(rx_errors, "rx_errors") \
1880 IFACE_STAT(tx_errors, "tx_errors") \
1881 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1882 IFACE_STAT(rx_over_errors, "rx_over_err") \
1883 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1884 IFACE_STAT(collisions, "collisions")
1886 #define IFACE_STAT(MEMBER, NAME) + 1
1887 enum { N_IFACE_STATS = IFACE_STATS };
1889 int64_t values[N_IFACE_STATS];
1890 char *keys[N_IFACE_STATS];
1893 struct netdev_stats stats;
1895 if (iface_is_synthetic(iface)) {
1899 /* Intentionally ignore return value, since errors will set 'stats' to
1900 * all-1s, and we will deal with that correctly below. */
1901 netdev_get_stats(iface->netdev, &stats);
1903 /* Copy statistics into keys[] and values[]. */
1905 #define IFACE_STAT(MEMBER, NAME) \
1906 if (stats.MEMBER != UINT64_MAX) { \
1908 values[n] = stats.MEMBER; \
1913 ovs_assert(n <= N_IFACE_STATS);
1915 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1920 br_refresh_stp_status(struct bridge *br)
1922 struct smap smap = SMAP_INITIALIZER(&smap);
1923 struct ofproto *ofproto = br->ofproto;
1924 struct ofproto_stp_status status;
1926 if (ofproto_get_stp_status(ofproto, &status)) {
1930 if (!status.enabled) {
1931 ovsrec_bridge_set_status(br->cfg, NULL);
1935 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1936 STP_ID_ARGS(status.bridge_id));
1937 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1938 STP_ID_ARGS(status.designated_root));
1939 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1941 ovsrec_bridge_set_status(br->cfg, &smap);
1942 smap_destroy(&smap);
1946 port_refresh_stp_status(struct port *port)
1948 struct ofproto *ofproto = port->bridge->ofproto;
1949 struct iface *iface;
1950 struct ofproto_port_stp_status status;
1953 if (port_is_synthetic(port)) {
1957 /* STP doesn't currently support bonds. */
1958 if (!list_is_singleton(&port->ifaces)) {
1959 ovsrec_port_set_status(port->cfg, NULL);
1963 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1964 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1968 if (!status.enabled) {
1969 ovsrec_port_set_status(port->cfg, NULL);
1973 /* Set Status column. */
1975 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1976 smap_add(&smap, "stp_state", stp_state_name(status.state));
1977 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1978 smap_add(&smap, "stp_role", stp_role_name(status.role));
1979 ovsrec_port_set_status(port->cfg, &smap);
1980 smap_destroy(&smap);
1984 port_refresh_stp_stats(struct port *port)
1986 struct ofproto *ofproto = port->bridge->ofproto;
1987 struct iface *iface;
1988 struct ofproto_port_stp_stats stats;
1990 int64_t int_values[3];
1992 if (port_is_synthetic(port)) {
1996 /* STP doesn't currently support bonds. */
1997 if (!list_is_singleton(&port->ifaces)) {
2001 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2002 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2006 if (!stats.enabled) {
2007 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2011 /* Set Statistics column. */
2012 keys[0] = "stp_tx_count";
2013 int_values[0] = stats.tx_count;
2014 keys[1] = "stp_rx_count";
2015 int_values[1] = stats.rx_count;
2016 keys[2] = "stp_error_count";
2017 int_values[2] = stats.error_count;
2019 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2020 ARRAY_SIZE(int_values));
2024 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2026 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2030 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2032 bool enable = enable_system_stats(cfg);
2034 system_stats_enable(enable);
2036 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2041 run_system_stats(void)
2043 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2046 stats = system_stats_run();
2048 struct ovsdb_idl_txn *txn;
2049 struct ovsdb_datum datum;
2051 txn = ovsdb_idl_txn_create(idl);
2052 ovsdb_datum_from_smap(&datum, stats);
2053 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2055 ovsdb_idl_txn_commit(txn);
2056 ovsdb_idl_txn_destroy(txn);
2063 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2066 case OFPCR12_ROLE_EQUAL:
2068 case OFPCR12_ROLE_MASTER:
2070 case OFPCR12_ROLE_SLAVE:
2072 case OFPCR12_ROLE_NOCHANGE:
2074 return "*** INVALID ROLE ***";
2079 refresh_controller_status(void)
2083 const struct ovsrec_controller *cfg;
2087 /* Accumulate status for controllers on all bridges. */
2088 HMAP_FOR_EACH (br, node, &all_bridges) {
2089 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2092 /* Update each controller in the database with current status. */
2093 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2094 struct ofproto_controller_info *cinfo =
2095 shash_find_data(&info, cfg->target);
2098 struct smap smap = SMAP_INITIALIZER(&smap);
2099 const char **values = cinfo->pairs.values;
2100 const char **keys = cinfo->pairs.keys;
2103 for (i = 0; i < cinfo->pairs.n; i++) {
2104 smap_add(&smap, keys[i], values[i]);
2107 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2108 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2110 ovsrec_controller_set_status(cfg, &smap);
2111 smap_destroy(&smap);
2113 ovsrec_controller_set_is_connected(cfg, false);
2114 ovsrec_controller_set_role(cfg, NULL);
2115 ovsrec_controller_set_status(cfg, NULL);
2119 ofproto_free_ofproto_controller_info(&info);
2124 * Some information in the database must be kept as up-to-date as possible to
2125 * allow controllers to respond rapidly to network outages. We call these
2126 * statistics "instant" stats.
2128 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2129 * assuming that they've changed. The only means we have to determine whether
2130 * they have changed are:
2132 * - Try to commit changes to the database. If nothing changed, then
2133 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2136 * - instant_stats_run() is called late in the run loop, after anything that
2137 * might change any of the instant stats.
2139 * We use these two facts together to avoid waking the process up every
2140 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2143 /* Minimum interval between writing updates to the instant stats to the
2145 #define INSTANT_INTERVAL_MSEC 100
2147 /* Current instant stats database transaction, NULL if there is no ongoing
2149 static struct ovsdb_idl_txn *instant_txn;
2151 /* Next time (in msec on monotonic clock) at which we will update the instant
2153 static long long int instant_next_txn = LLONG_MIN;
2155 /* True if the run loop has run since we last saw that the instant stats were
2156 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2157 * to refresh the instant stats. */
2158 static bool instant_stats_could_have_changed;
2161 instant_stats_run(void)
2163 enum ovsdb_idl_txn_status status;
2165 instant_stats_could_have_changed = true;
2171 if (time_msec() < instant_next_txn) {
2174 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2176 seq = seq_read(connectivity_seq_get());
2177 if (seq == connectivity_seqno) {
2180 connectivity_seqno = seq;
2182 instant_txn = ovsdb_idl_txn_create(idl);
2183 HMAP_FOR_EACH (br, node, &all_bridges) {
2184 struct iface *iface;
2187 br_refresh_stp_status(br);
2189 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2190 port_refresh_stp_status(port);
2193 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2194 enum netdev_flags flags;
2196 const char *link_state;
2197 int64_t link_resets;
2200 if (iface_is_synthetic(iface)) {
2204 current = ofproto_port_is_lacp_current(br->ofproto,
2208 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2210 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2213 error = netdev_get_flags(iface->netdev, &flags);
2215 const char *state = flags & NETDEV_UP ? "up" : "down";
2216 ovsrec_interface_set_admin_state(iface->cfg, state);
2218 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2221 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2222 ovsrec_interface_set_link_state(iface->cfg, link_state);
2224 link_resets = netdev_get_carrier_resets(iface->netdev);
2225 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2227 iface_refresh_cfm_stats(iface);
2230 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2232 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2233 smap_destroy(&smap);
2238 status = ovsdb_idl_txn_commit(instant_txn);
2239 if (status != TXN_INCOMPLETE) {
2240 ovsdb_idl_txn_destroy(instant_txn);
2243 if (status == TXN_UNCHANGED) {
2244 instant_stats_could_have_changed = false;
2249 instant_stats_wait(void)
2252 ovsdb_idl_txn_wait(instant_txn);
2253 } else if (instant_stats_could_have_changed) {
2254 poll_timer_wait_until(instant_next_txn);
2265 /* Let each datapath type do the work that it needs to do. */
2267 ofproto_enumerate_types(&types);
2268 SSET_FOR_EACH (type, &types) {
2269 ofproto_type_run(type);
2271 sset_destroy(&types);
2273 /* Let each bridge do the work that it needs to do. */
2274 HMAP_FOR_EACH (br, node, &all_bridges) {
2275 ofproto_run(br->ofproto);
2282 static struct ovsrec_open_vswitch null_cfg;
2283 const struct ovsrec_open_vswitch *cfg;
2285 bool vlan_splinters_changed;
2288 ovsrec_open_vswitch_init(&null_cfg);
2292 if (ovsdb_idl_is_lock_contended(idl)) {
2293 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2294 struct bridge *br, *next_br;
2296 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2297 "disabling this process (pid %ld) until it goes away",
2298 (long int) getpid());
2300 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2303 /* Since we will not be running system_stats_run() in this process
2304 * with the current situation of multiple ovs-vswitchd daemons,
2305 * disable system stats collection. */
2306 system_stats_enable(false);
2308 } else if (!ovsdb_idl_has_lock(idl)) {
2311 cfg = ovsrec_open_vswitch_first(idl);
2313 /* Initialize the ofproto library. This only needs to run once, but
2314 * it must be done after the configuration is set. If the
2315 * initialization has already occurred, bridge_init_ofproto()
2316 * returns immediately. */
2317 bridge_init_ofproto(cfg);
2319 /* Once the value of flow-restore-wait is false, we no longer should
2320 * check its value from the database. */
2321 if (cfg && ofproto_get_flow_restore_wait()) {
2322 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2323 "flow-restore-wait", false));
2328 /* Re-configure SSL. We do this on every trip through the main loop,
2329 * instead of just when the database changes, because the contents of the
2330 * key and certificate files can change without the database changing.
2332 * We do this before bridge_reconfigure() because that function might
2333 * initiate SSL connections and thus requires SSL to be configured. */
2334 if (cfg && cfg->ssl) {
2335 const struct ovsrec_ssl *ssl = cfg->ssl;
2337 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2338 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2341 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2342 * usage has changed. */
2343 vlan_splinters_changed = false;
2344 if (vlan_splinters_enabled_anywhere) {
2345 HMAP_FOR_EACH (br, node, &all_bridges) {
2346 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2347 vlan_splinters_changed = true;
2353 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2354 struct ovsdb_idl_txn *txn;
2356 idl_seqno = ovsdb_idl_get_seqno(idl);
2357 txn = ovsdb_idl_txn_create(idl);
2358 bridge_reconfigure(cfg ? cfg : &null_cfg);
2361 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2364 /* If we are completing our initial configuration for this run
2365 * of ovs-vswitchd, then keep the transaction around to monitor
2366 * it for completion. */
2367 if (initial_config_done) {
2368 ovsdb_idl_txn_commit(txn);
2369 ovsdb_idl_txn_destroy(txn);
2371 initial_config_done = true;
2372 daemonize_txn = txn;
2376 if (daemonize_txn) {
2377 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2378 if (status != TXN_INCOMPLETE) {
2379 ovsdb_idl_txn_destroy(daemonize_txn);
2380 daemonize_txn = NULL;
2382 /* ovs-vswitchd has completed initialization, so allow the
2383 * process that forked us to exit successfully. */
2384 daemonize_complete();
2386 vlog_enable_async();
2388 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2392 /* Refresh interface and mirror stats if necessary. */
2393 if (time_msec() >= iface_stats_timer) {
2395 struct ovsdb_idl_txn *txn;
2397 txn = ovsdb_idl_txn_create(idl);
2398 HMAP_FOR_EACH (br, node, &all_bridges) {
2402 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2403 struct iface *iface;
2405 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2406 iface_refresh_stats(iface);
2407 iface_refresh_status(iface);
2410 port_refresh_stp_stats(port);
2413 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2414 mirror_refresh_stats(m);
2418 refresh_controller_status();
2419 ovsdb_idl_txn_commit(txn);
2420 ovsdb_idl_txn_destroy(txn); /* XXX */
2423 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2427 instant_stats_run();
2436 ovsdb_idl_wait(idl);
2437 if (daemonize_txn) {
2438 ovsdb_idl_txn_wait(daemonize_txn);
2442 ofproto_enumerate_types(&types);
2443 SSET_FOR_EACH (type, &types) {
2444 ofproto_type_wait(type);
2446 sset_destroy(&types);
2448 if (!hmap_is_empty(&all_bridges)) {
2451 HMAP_FOR_EACH (br, node, &all_bridges) {
2452 ofproto_wait(br->ofproto);
2454 poll_timer_wait_until(iface_stats_timer);
2457 system_stats_wait();
2458 instant_stats_wait();
2461 /* Adds some memory usage statistics for bridges into 'usage', for use with
2462 * memory_report(). */
2464 bridge_get_memory_usage(struct simap *usage)
2471 ofproto_enumerate_types(&types);
2472 SSET_FOR_EACH (type, &types) {
2473 ofproto_type_get_memory_usage(type, usage);
2475 sset_destroy(&types);
2477 HMAP_FOR_EACH (br, node, &all_bridges) {
2478 ofproto_get_memory_usage(br->ofproto, usage);
2482 /* QoS unixctl user interface functions. */
2484 struct qos_unixctl_show_cbdata {
2486 struct iface *iface;
2490 qos_unixctl_show_queue(unsigned int queue_id,
2491 const struct smap *details,
2492 struct iface *iface,
2495 struct netdev_queue_stats stats;
2496 struct smap_node *node;
2499 ds_put_cstr(ds, "\n");
2501 ds_put_format(ds, "Queue %u:\n", queue_id);
2503 ds_put_cstr(ds, "Default:\n");
2506 SMAP_FOR_EACH (node, details) {
2507 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2510 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2512 if (stats.tx_packets != UINT64_MAX) {
2513 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2516 if (stats.tx_bytes != UINT64_MAX) {
2517 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2520 if (stats.tx_errors != UINT64_MAX) {
2521 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2524 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2525 queue_id, ovs_strerror(error));
2530 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2531 const char *argv[], void *aux OVS_UNUSED)
2533 struct ds ds = DS_EMPTY_INITIALIZER;
2534 struct smap smap = SMAP_INITIALIZER(&smap);
2535 struct iface *iface;
2537 struct smap_node *node;
2539 iface = iface_find(argv[1]);
2541 unixctl_command_reply_error(conn, "no such interface");
2545 netdev_get_qos(iface->netdev, &type, &smap);
2547 if (*type != '\0') {
2548 struct netdev_queue_dump dump;
2549 struct smap details;
2550 unsigned int queue_id;
2552 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2554 SMAP_FOR_EACH (node, &smap) {
2555 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2558 smap_init(&details);
2559 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2560 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2562 smap_destroy(&details);
2564 unixctl_command_reply(conn, ds_cstr(&ds));
2566 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2567 unixctl_command_reply_error(conn, ds_cstr(&ds));
2570 smap_destroy(&smap);
2574 /* Bridge reconfiguration functions. */
2576 bridge_create(const struct ovsrec_bridge *br_cfg)
2580 ovs_assert(!bridge_lookup(br_cfg->name));
2581 br = xzalloc(sizeof *br);
2583 br->name = xstrdup(br_cfg->name);
2584 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2587 /* Derive the default Ethernet address from the bridge's UUID. This should
2588 * be unique and it will be stable between ovs-vswitchd runs. */
2589 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2590 eth_addr_mark_random(br->default_ea);
2592 hmap_init(&br->ports);
2593 hmap_init(&br->ifaces);
2594 hmap_init(&br->iface_by_name);
2595 hmap_init(&br->mirrors);
2597 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2601 bridge_destroy(struct bridge *br)
2604 struct mirror *mirror, *next_mirror;
2605 struct port *port, *next_port;
2607 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2610 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2611 mirror_destroy(mirror);
2614 hmap_remove(&all_bridges, &br->node);
2615 ofproto_destroy(br->ofproto);
2616 hmap_destroy(&br->ifaces);
2617 hmap_destroy(&br->ports);
2618 hmap_destroy(&br->iface_by_name);
2619 hmap_destroy(&br->mirrors);
2626 static struct bridge *
2627 bridge_lookup(const char *name)
2631 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2632 if (!strcmp(br->name, name)) {
2639 /* Handle requests for a listing of all flows known by the OpenFlow
2640 * stack, including those normally hidden. */
2642 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2643 const char *argv[], void *aux OVS_UNUSED)
2648 br = bridge_lookup(argv[1]);
2650 unixctl_command_reply_error(conn, "Unknown bridge");
2655 ofproto_get_all_flows(br->ofproto, &results);
2657 unixctl_command_reply(conn, ds_cstr(&results));
2658 ds_destroy(&results);
2661 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2662 * connections and reconnect. If BRIDGE is not specified, then all bridges
2663 * drop their controller connections and reconnect. */
2665 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2666 const char *argv[], void *aux OVS_UNUSED)
2670 br = bridge_lookup(argv[1]);
2672 unixctl_command_reply_error(conn, "Unknown bridge");
2675 ofproto_reconnect_controllers(br->ofproto);
2677 HMAP_FOR_EACH (br, node, &all_bridges) {
2678 ofproto_reconnect_controllers(br->ofproto);
2681 unixctl_command_reply(conn, NULL);
2685 bridge_get_controllers(const struct bridge *br,
2686 struct ovsrec_controller ***controllersp)
2688 struct ovsrec_controller **controllers;
2689 size_t n_controllers;
2691 controllers = br->cfg->controller;
2692 n_controllers = br->cfg->n_controller;
2694 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2700 *controllersp = controllers;
2702 return n_controllers;
2706 bridge_collect_wanted_ports(struct bridge *br,
2707 const unsigned long int *splinter_vlans,
2708 struct shash *wanted_ports)
2712 shash_init(wanted_ports);
2714 for (i = 0; i < br->cfg->n_ports; i++) {
2715 const char *name = br->cfg->ports[i]->name;
2716 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2717 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2721 if (bridge_get_controllers(br, NULL)
2722 && !shash_find(wanted_ports, br->name)) {
2723 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2724 br->name, br->name);
2726 ovsrec_interface_init(&br->synth_local_iface);
2727 ovsrec_port_init(&br->synth_local_port);
2729 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2730 br->synth_local_port.n_interfaces = 1;
2731 br->synth_local_port.name = br->name;
2733 br->synth_local_iface.name = br->name;
2734 br->synth_local_iface.type = "internal";
2736 br->synth_local_ifacep = &br->synth_local_iface;
2738 shash_add(wanted_ports, br->name, &br->synth_local_port);
2741 if (splinter_vlans) {
2742 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2746 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2747 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2748 * 'br' needs to complete its configuration. */
2750 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2752 struct shash_node *port_node;
2753 struct port *port, *next;
2755 /* Get rid of deleted ports.
2756 * Get rid of deleted interfaces on ports that still exist. */
2757 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2758 port->cfg = shash_find_data(wanted_ports, port->name);
2762 port_del_ifaces(port);
2766 /* Update iface->cfg and iface->type in interfaces that still exist. */
2767 SHASH_FOR_EACH (port_node, wanted_ports) {
2768 const struct ovsrec_port *port = port_node->data;
2771 for (i = 0; i < port->n_interfaces; i++) {
2772 const struct ovsrec_interface *cfg = port->interfaces[i];
2773 struct iface *iface = iface_lookup(br, cfg->name);
2774 const char *type = iface_get_type(cfg, br->cfg);
2779 } else if (!strcmp(type, "null")) {
2780 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2781 " may be removed in February 2013. Please email"
2782 " dev@openvswitch.org with concerns.",
2785 /* We will add new interfaces later. */
2791 /* Initializes 'oc' appropriately as a management service controller for
2794 * The caller must free oc->target when it is no longer needed. */
2796 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2797 struct ofproto_controller *oc)
2799 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2800 oc->max_backoff = 0;
2801 oc->probe_interval = 60;
2802 oc->band = OFPROTO_OUT_OF_BAND;
2804 oc->burst_limit = 0;
2805 oc->enable_async_msgs = true;
2808 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2810 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2811 struct ofproto_controller *oc)
2815 oc->target = c->target;
2816 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2817 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2818 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2819 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2820 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2821 oc->burst_limit = (c->controller_burst_limit
2822 ? *c->controller_burst_limit : 0);
2823 oc->enable_async_msgs = (!c->enable_async_messages
2824 || *c->enable_async_messages);
2825 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2826 if (dscp < 0 || dscp > 63) {
2827 dscp = DSCP_DEFAULT;
2832 /* Configures the IP stack for 'br''s local interface properly according to the
2833 * configuration in 'c'. */
2835 bridge_configure_local_iface_netdev(struct bridge *br,
2836 struct ovsrec_controller *c)
2838 struct netdev *netdev;
2839 struct in_addr mask, gateway;
2841 struct iface *local_iface;
2844 /* If there's no local interface or no IP address, give up. */
2845 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2846 if (!local_iface || !c->local_ip
2847 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2851 /* Bring up the local interface. */
2852 netdev = local_iface->netdev;
2853 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2855 /* Configure the IP address and netmask. */
2856 if (!c->local_netmask
2857 || !inet_pton(AF_INET, c->local_netmask, &mask)
2859 mask.s_addr = guess_netmask(ip.s_addr);
2861 if (!netdev_set_in4(netdev, ip, mask)) {
2862 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2863 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2866 /* Configure the default gateway. */
2867 if (c->local_gateway
2868 && inet_pton(AF_INET, c->local_gateway, &gateway)
2869 && gateway.s_addr) {
2870 if (!netdev_add_router(netdev, gateway)) {
2871 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2872 br->name, IP_ARGS(gateway.s_addr));
2877 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2878 * in either string are treated as equal to any number of slashes in the other,
2879 * e.g. "x///y" is equal to "x/y".
2881 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2882 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2883 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2884 * 'b' against a prefix of 'a'.
2887 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2889 const char *b_start = b;
2891 if (b - b_start >= b_stoplen) {
2893 } else if (*a != *b) {
2895 } else if (*a == '/') {
2896 a += strspn(a, "/");
2897 b += strspn(b, "/");
2898 } else if (*a == '\0') {
2908 bridge_configure_remotes(struct bridge *br,
2909 const struct sockaddr_in *managers, size_t n_managers)
2911 bool disable_in_band;
2913 struct ovsrec_controller **controllers;
2914 size_t n_controllers;
2916 enum ofproto_fail_mode fail_mode;
2918 struct ofproto_controller *ocs;
2922 /* Check if we should disable in-band control on this bridge. */
2923 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2926 /* Set OpenFlow queue ID for in-band control. */
2927 ofproto_set_in_band_queue(br->ofproto,
2928 smap_get_int(&br->cfg->other_config,
2929 "in-band-queue", -1));
2931 if (disable_in_band) {
2932 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2934 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2937 n_controllers = bridge_get_controllers(br, &controllers);
2939 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2942 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2943 for (i = 0; i < n_controllers; i++) {
2944 struct ovsrec_controller *c = controllers[i];
2946 if (!strncmp(c->target, "punix:", 6)
2947 || !strncmp(c->target, "unix:", 5)) {
2948 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2951 if (!strncmp(c->target, "unix:", 5)) {
2952 /* Connect to a listening socket */
2953 whitelist = xasprintf("unix:%s/", ovs_rundir());
2954 if (strchr(c->target, '/') &&
2955 !equal_pathnames(c->target, whitelist,
2956 strlen(whitelist))) {
2957 /* Absolute path specified, but not in ovs_rundir */
2958 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2959 "controller \"%s\" due to possibility for "
2960 "remote exploit. Instead, specify socket "
2961 "in whitelisted \"%s\" or connect to "
2962 "\"unix:%s/%s.mgmt\" (which is always "
2963 "available without special configuration).",
2964 br->name, c->target, whitelist,
2965 ovs_rundir(), br->name);
2970 whitelist = xasprintf("punix:%s/%s.controller",
2971 ovs_rundir(), br->name);
2972 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2973 /* Prevent remote ovsdb-server users from accessing
2974 * arbitrary Unix domain sockets and overwriting arbitrary
2976 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2977 "controller \"%s\" due to possibility of "
2978 "overwriting local files. Instead, specify "
2979 "whitelisted \"%s\" or connect to "
2980 "\"unix:%s/%s.mgmt\" (which is always "
2981 "available without special configuration).",
2982 br->name, c->target, whitelist,
2983 ovs_rundir(), br->name);
2992 bridge_configure_local_iface_netdev(br, c);
2993 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2994 if (disable_in_band) {
2995 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3000 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3001 bridge_get_allowed_versions(br));
3002 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3005 /* Set the fail-mode. */
3006 fail_mode = !br->cfg->fail_mode
3007 || !strcmp(br->cfg->fail_mode, "standalone")
3008 ? OFPROTO_FAIL_STANDALONE
3009 : OFPROTO_FAIL_SECURE;
3010 ofproto_set_fail_mode(br->ofproto, fail_mode);
3012 /* Configure OpenFlow controller connection snooping. */
3013 if (!ofproto_has_snoops(br->ofproto)) {
3017 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3018 ovs_rundir(), br->name));
3019 ofproto_set_snoops(br->ofproto, &snoops);
3020 sset_destroy(&snoops);
3025 bridge_configure_tables(struct bridge *br)
3027 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3031 n_tables = ofproto_get_n_tables(br->ofproto);
3033 for (i = 0; i < n_tables; i++) {
3034 struct ofproto_table_settings s;
3037 s.max_flows = UINT_MAX;
3040 s.n_prefix_fields = 0;
3041 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3043 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3044 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3047 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3048 s.max_flows = *cfg->flow_limit;
3050 if (cfg->overflow_policy
3051 && !strcmp(cfg->overflow_policy, "evict")) {
3053 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3054 for (k = 0; k < cfg->n_groups; k++) {
3055 const char *string = cfg->groups[k];
3058 msg = mf_parse_subfield__(&s.groups[k], &string);
3060 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3061 "'groups' (%s)", br->name, i, msg);
3063 } else if (*string) {
3064 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3065 "element '%s' contains trailing garbage",
3066 br->name, i, cfg->groups[k]);
3072 /* Prefix lookup fields. */
3073 s.n_prefix_fields = 0;
3074 for (k = 0; k < cfg->n_prefixes; k++) {
3075 const char *name = cfg->prefixes[k];
3076 const struct mf_field *mf = mf_from_name(name);
3078 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3082 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3083 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3084 "%s", br->name, name);
3087 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3088 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3089 "field not used: %s", br->name, name);
3092 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3094 if (s.n_prefix_fields > 0) {
3096 struct ds ds = DS_EMPTY_INITIALIZER;
3097 for (k = 0; k < s.n_prefix_fields; k++) {
3099 ds_put_char(&ds, ',');
3101 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3103 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3104 br->name, i, ds_cstr(&ds));
3109 ofproto_configure_table(br->ofproto, i, &s);
3113 for (; j < br->cfg->n_flow_tables; j++) {
3114 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3115 "%"PRId64" not supported by this datapath", br->name,
3116 br->cfg->key_flow_tables[j]);
3121 bridge_configure_dp_desc(struct bridge *br)
3123 ofproto_set_dp_desc(br->ofproto,
3124 smap_get(&br->cfg->other_config, "dp-desc"));
3127 /* Port functions. */
3129 static void iface_destroy__(struct iface *);
3131 static struct port *
3132 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3136 port = xzalloc(sizeof *port);
3138 port->name = xstrdup(cfg->name);
3140 list_init(&port->ifaces);
3142 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3146 /* Deletes interfaces from 'port' that are no longer configured for it. */
3148 port_del_ifaces(struct port *port)
3150 struct iface *iface, *next;
3151 struct sset new_ifaces;
3154 /* Collect list of new interfaces. */
3155 sset_init(&new_ifaces);
3156 for (i = 0; i < port->cfg->n_interfaces; i++) {
3157 const char *name = port->cfg->interfaces[i]->name;
3158 const char *type = port->cfg->interfaces[i]->type;
3159 if (strcmp(type, "null")) {
3160 sset_add(&new_ifaces, name);
3164 /* Get rid of deleted interfaces. */
3165 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3166 if (!sset_contains(&new_ifaces, iface->name)) {
3167 iface_destroy(iface);
3171 sset_destroy(&new_ifaces);
3175 port_destroy(struct port *port)
3178 struct bridge *br = port->bridge;
3179 struct iface *iface, *next;
3182 ofproto_bundle_unregister(br->ofproto, port);
3185 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3186 iface_destroy__(iface);
3189 hmap_remove(&br->ports, &port->hmap_node);
3195 static struct port *
3196 port_lookup(const struct bridge *br, const char *name)
3200 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3202 if (!strcmp(port->name, name)) {
3210 enable_lacp(struct port *port, bool *activep)
3212 if (!port->cfg->lacp) {
3213 /* XXX when LACP implementation has been sufficiently tested, enable by
3214 * default and make active on bonded ports. */
3216 } else if (!strcmp(port->cfg->lacp, "off")) {
3218 } else if (!strcmp(port->cfg->lacp, "active")) {
3221 } else if (!strcmp(port->cfg->lacp, "passive")) {
3225 VLOG_WARN("port %s: unknown LACP mode %s",
3226 port->name, port->cfg->lacp);
3231 static struct lacp_settings *
3232 port_configure_lacp(struct port *port, struct lacp_settings *s)
3234 const char *lacp_time, *system_id;
3237 if (!enable_lacp(port, &s->active)) {
3241 s->name = port->name;
3243 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3245 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3246 ETH_ADDR_SCAN_ARGS(s->id))) {
3247 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3248 " address.", port->name, system_id);
3252 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3255 if (eth_addr_is_zero(s->id)) {
3256 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3260 /* Prefer bondable links if unspecified. */
3261 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3263 s->priority = (priority > 0 && priority <= UINT16_MAX
3265 : UINT16_MAX - !list_is_short(&port->ifaces));
3267 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3268 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3270 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3271 "lacp-fallback-ab", false);
3277 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3279 int priority, portid, key;
3281 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3282 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3284 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3286 if (portid <= 0 || portid > UINT16_MAX) {
3287 portid = ofp_to_u16(iface->ofp_port);
3290 if (priority <= 0 || priority > UINT16_MAX) {
3291 priority = UINT16_MAX;
3294 if (key < 0 || key > UINT16_MAX) {
3298 s->name = iface->name;
3300 s->priority = priority;
3305 port_configure_bond(struct port *port, struct bond_settings *s)
3307 const char *detect_s;
3308 struct iface *iface;
3309 int miimon_interval;
3311 s->name = port->name;
3313 if (port->cfg->bond_mode) {
3314 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3315 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3316 port->name, port->cfg->bond_mode,
3317 bond_mode_to_string(s->balance));
3320 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3322 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3323 * active-backup. At some point we should remove this warning. */
3324 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3325 " in previous versions, the default bond_mode was"
3326 " balance-slb", port->name,
3327 bond_mode_to_string(s->balance));
3329 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3330 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3331 "please use another bond type or disable flood_vlans",
3335 miimon_interval = smap_get_int(&port->cfg->other_config,
3336 "bond-miimon-interval", 0);
3337 if (miimon_interval <= 0) {
3338 miimon_interval = 200;
3341 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3342 if (!detect_s || !strcmp(detect_s, "carrier")) {
3343 miimon_interval = 0;
3344 } else if (strcmp(detect_s, "miimon")) {
3345 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3346 "defaulting to carrier", port->name, detect_s);
3347 miimon_interval = 0;
3350 s->up_delay = MAX(0, port->cfg->bond_updelay);
3351 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3352 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3353 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3354 "bond-rebalance-interval", 10000);
3355 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3356 s->rebalance_interval = 1000;
3359 s->fake_iface = port->cfg->bond_fake_iface;
3361 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3362 "lacp-fallback-ab", false);
3364 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3365 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3369 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3370 * instead of obtaining it from the database. */
3372 port_is_synthetic(const struct port *port)
3374 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3377 /* Interface functions. */
3380 iface_is_internal(const struct ovsrec_interface *iface,
3381 const struct ovsrec_bridge *br)
3383 /* The local port and "internal" ports are always "internal". */
3384 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3387 /* Returns the correct network device type for interface 'iface' in bridge
3390 iface_get_type(const struct ovsrec_interface *iface,
3391 const struct ovsrec_bridge *br)
3395 /* The local port always has type "internal". Other ports take
3396 * their type from the database and default to "system" if none is
3398 if (iface_is_internal(iface, br)) {
3401 type = iface->type[0] ? iface->type : "system";
3404 return ofproto_port_open_type(br->datapath_type, type);
3408 iface_destroy__(struct iface *iface)
3411 struct port *port = iface->port;
3412 struct bridge *br = port->bridge;
3414 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3415 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3418 if (iface->ofp_port != OFPP_NONE) {
3419 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3422 list_remove(&iface->port_elem);
3423 hmap_remove(&br->iface_by_name, &iface->name_node);
3425 netdev_close(iface->netdev);
3433 iface_destroy(struct iface *iface)
3436 struct port *port = iface->port;
3438 iface_destroy__(iface);
3439 if (list_is_empty(&port->ifaces)) {
3445 static struct iface *
3446 iface_lookup(const struct bridge *br, const char *name)
3448 struct iface *iface;
3450 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3451 &br->iface_by_name) {
3452 if (!strcmp(iface->name, name)) {
3460 static struct iface *
3461 iface_find(const char *name)
3463 const struct bridge *br;
3465 HMAP_FOR_EACH (br, node, &all_bridges) {
3466 struct iface *iface = iface_lookup(br, name);
3475 static struct iface *
3476 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3478 struct iface *iface;
3480 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3482 if (iface->ofp_port == ofp_port) {
3489 /* Set Ethernet address of 'iface', if one is specified in the configuration
3492 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3494 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3495 struct iface *hw_addr_iface;
3497 if (strcmp(iface->type, "internal")) {
3501 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3503 } else if (port->cfg->fake_bridge) {
3504 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3505 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3510 if (iface->ofp_port == OFPP_LOCAL) {
3511 VLOG_ERR("interface %s: ignoring mac in Interface record "
3512 "(use Bridge record to set local port's mac)",
3514 } else if (eth_addr_is_multicast(mac)) {
3515 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3518 int error = netdev_set_etheraddr(iface->netdev, mac);
3520 VLOG_ERR("interface %s: setting MAC failed (%s)",
3521 iface->name, ovs_strerror(error));
3527 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3529 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3531 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3532 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3533 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3537 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3538 * sets the "ofport" field to -1.
3540 * This is appropriate when 'if_cfg''s interface cannot be created or is
3541 * otherwise invalid. */
3543 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3545 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3546 iface_set_ofport(if_cfg, OFPP_NONE);
3547 ovsrec_interface_set_status(if_cfg, NULL);
3548 ovsrec_interface_set_admin_state(if_cfg, NULL);
3549 ovsrec_interface_set_duplex(if_cfg, NULL);
3550 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3551 ovsrec_interface_set_link_state(if_cfg, NULL);
3552 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3553 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3554 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3555 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3556 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3557 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3558 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3559 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3564 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3566 union ovsdb_atom atom;
3568 atom.integer = target;
3569 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3573 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3575 struct ofpbuf queues_buf;
3577 ofpbuf_init(&queues_buf, 0);
3579 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3580 netdev_set_qos(iface->netdev, NULL, NULL);
3582 const struct ovsdb_datum *queues;
3583 struct netdev_queue_dump dump;
3584 unsigned int queue_id;
3585 struct smap details;
3589 /* Configure top-level Qos for 'iface'. */
3590 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3592 /* Deconfigure queues that were deleted. */
3593 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3595 smap_init(&details);
3596 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3597 if (!queue_ids_include(queues, queue_id)) {
3598 netdev_delete_queue(iface->netdev, queue_id);
3601 smap_destroy(&details);
3603 /* Configure queues for 'iface'. */
3605 for (i = 0; i < qos->n_queues; i++) {
3606 const struct ovsrec_queue *queue = qos->value_queues[i];
3607 unsigned int queue_id = qos->key_queues[i];
3609 if (queue_id == 0) {
3613 if (queue->n_dscp == 1) {
3614 struct ofproto_port_queue *port_queue;
3616 port_queue = ofpbuf_put_uninit(&queues_buf,
3617 sizeof *port_queue);
3618 port_queue->queue = queue_id;
3619 port_queue->dscp = queue->dscp[0];
3622 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3625 struct smap details;
3627 smap_init(&details);
3628 netdev_set_queue(iface->netdev, 0, &details);
3629 smap_destroy(&details);
3633 if (iface->ofp_port != OFPP_NONE) {
3634 const struct ofproto_port_queue *port_queues = queues_buf.data;
3635 size_t n_queues = queues_buf.size / sizeof *port_queues;
3637 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3638 port_queues, n_queues);
3641 netdev_set_policing(iface->netdev,
3642 iface->cfg->ingress_policing_rate,
3643 iface->cfg->ingress_policing_burst);
3645 ofpbuf_uninit(&queues_buf);
3649 iface_configure_cfm(struct iface *iface)
3651 const struct ovsrec_interface *cfg = iface->cfg;
3652 const char *opstate_str;
3653 const char *cfm_ccm_vlan;
3654 struct cfm_settings s;
3655 struct smap netdev_args;
3657 if (!cfg->n_cfm_mpid) {
3658 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3662 s.check_tnl_key = false;
3663 smap_init(&netdev_args);
3664 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3665 const char *key = smap_get(&netdev_args, "key");
3666 const char *in_key = smap_get(&netdev_args, "in_key");
3668 s.check_tnl_key = (key && !strcmp(key, "flow"))
3669 || (in_key && !strcmp(in_key, "flow"));
3671 smap_destroy(&netdev_args);
3673 s.mpid = *cfg->cfm_mpid;
3674 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3675 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3676 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3678 if (s.interval <= 0) {
3682 if (!cfm_ccm_vlan) {
3684 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3685 s.ccm_vlan = CFM_RANDOM_VLAN;
3687 s.ccm_vlan = atoi(cfm_ccm_vlan);
3688 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3693 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3695 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3697 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3698 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3700 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3703 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3704 * instead of obtaining it from the database. */
3706 iface_is_synthetic(const struct iface *iface)
3708 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3712 iface_validate_ofport__(size_t n, int64_t *ofport)
3714 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3715 ? u16_to_ofp(*ofport)
3720 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3722 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3726 iface_pick_ofport(const struct ovsrec_interface *cfg)
3728 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3729 return (requested_ofport != OFPP_NONE
3731 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3734 /* Port mirroring. */
3736 static struct mirror *
3737 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3741 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3742 if (uuid_equals(uuid, &m->uuid)) {
3750 bridge_configure_mirrors(struct bridge *br)
3752 const struct ovsdb_datum *mc;
3753 unsigned long *flood_vlans;
3754 struct mirror *m, *next;
3757 /* Get rid of deleted mirrors. */
3758 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3759 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3760 union ovsdb_atom atom;
3762 atom.uuid = m->uuid;
3763 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3768 /* Add new mirrors and reconfigure existing ones. */
3769 for (i = 0; i < br->cfg->n_mirrors; i++) {
3770 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3771 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3773 m = mirror_create(br, cfg);
3776 if (!mirror_configure(m)) {
3781 /* Update flooded vlans (for RSPAN). */
3782 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3783 br->cfg->n_flood_vlans);
3784 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3785 bitmap_free(flood_vlans);
3788 static struct mirror *
3789 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3793 m = xzalloc(sizeof *m);
3794 m->uuid = cfg->header_.uuid;
3795 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3797 m->name = xstrdup(cfg->name);
3803 mirror_destroy(struct mirror *m)
3806 struct bridge *br = m->bridge;
3809 ofproto_mirror_unregister(br->ofproto, m);
3812 hmap_remove(&br->mirrors, &m->hmap_node);
3819 mirror_collect_ports(struct mirror *m,
3820 struct ovsrec_port **in_ports, int n_in_ports,
3821 void ***out_portsp, size_t *n_out_portsp)
3823 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3824 size_t n_out_ports = 0;
3827 for (i = 0; i < n_in_ports; i++) {
3828 const char *name = in_ports[i]->name;
3829 struct port *port = port_lookup(m->bridge, name);
3831 out_ports[n_out_ports++] = port;
3833 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3834 "port %s", m->bridge->name, m->name, name);
3837 *out_portsp = out_ports;
3838 *n_out_portsp = n_out_ports;
3842 mirror_configure(struct mirror *m)
3844 const struct ovsrec_mirror *cfg = m->cfg;
3845 struct ofproto_mirror_settings s;
3848 if (strcmp(cfg->name, m->name)) {
3850 m->name = xstrdup(cfg->name);
3854 /* Get output port or VLAN. */
3855 if (cfg->output_port) {
3856 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3857 if (!s.out_bundle) {
3858 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3859 m->bridge->name, m->name);
3862 s.out_vlan = UINT16_MAX;
3864 if (cfg->output_vlan) {
3865 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3866 "output vlan; ignoring output vlan",
3867 m->bridge->name, m->name);
3869 } else if (cfg->output_vlan) {
3870 /* The database should prevent invalid VLAN values. */
3871 s.out_bundle = NULL;
3872 s.out_vlan = *cfg->output_vlan;
3874 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3875 m->bridge->name, m->name);
3879 /* Get port selection. */
3880 if (cfg->select_all) {
3881 size_t n_ports = hmap_count(&m->bridge->ports);
3882 void **ports = xmalloc(n_ports * sizeof *ports);
3887 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3897 /* Get ports, dropping ports that don't exist.
3898 * The IDL ensures that there are no duplicates. */
3899 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3900 &s.srcs, &s.n_srcs);
3901 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3902 &s.dsts, &s.n_dsts);
3905 /* Get VLAN selection. */
3906 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3909 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3912 if (s.srcs != s.dsts) {
3921 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3923 * This is deprecated. It is only for compatibility with broken device drivers
3924 * in old versions of Linux that do not properly support VLANs when VLAN
3925 * devices are not used. When broken device drivers are no longer in
3926 * widespread use, we will delete these interfaces. */
3928 static struct ovsrec_port **recs;
3929 static size_t n_recs, allocated_recs;
3931 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3932 * splinters are reconfigured. */
3934 register_rec(struct ovsrec_port *rec)
3936 if (n_recs >= allocated_recs) {
3937 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3939 recs[n_recs++] = rec;
3942 /* Frees all of the ports registered with register_reg(). */
3944 free_registered_recs(void)
3948 for (i = 0; i < n_recs; i++) {
3949 struct ovsrec_port *port = recs[i];
3952 for (j = 0; j < port->n_interfaces; j++) {
3953 struct ovsrec_interface *iface = port->interfaces[j];
3958 smap_destroy(&port->other_config);
3959 free(port->interfaces);
3967 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3970 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3972 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3976 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3979 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3980 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3981 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3984 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3985 * use, returns NULL.
3987 * Updates 'vlan_splinters_enabled_anywhere'. */
3988 static unsigned long int *
3989 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3991 unsigned long int *splinter_vlans;
3992 struct sset splinter_ifaces;
3993 const char *real_dev_name;
3994 struct shash *real_devs;
3995 struct shash_node *node;
3999 /* Free space allocated for synthesized ports and interfaces, since we're
4000 * in the process of reconstructing all of them. */
4001 free_registered_recs();
4003 splinter_vlans = bitmap_allocate(4096);
4004 sset_init(&splinter_ifaces);
4005 vlan_splinters_enabled_anywhere = false;
4006 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4007 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4010 for (j = 0; j < br_cfg->n_ports; j++) {
4011 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4014 for (k = 0; k < port_cfg->n_interfaces; k++) {
4015 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4017 if (vlan_splinters_is_enabled(iface_cfg)) {
4018 vlan_splinters_enabled_anywhere = true;
4019 sset_add(&splinter_ifaces, iface_cfg->name);
4020 vlan_bitmap_from_array__(port_cfg->trunks,
4026 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4027 bitmap_set1(splinter_vlans, *port_cfg->tag);
4032 if (!vlan_splinters_enabled_anywhere) {
4033 free(splinter_vlans);
4034 sset_destroy(&splinter_ifaces);
4038 HMAP_FOR_EACH (br, node, &all_bridges) {
4040 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4044 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4045 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4046 * device to be created for it. */
4047 bitmap_set0(splinter_vlans, 0);
4048 bitmap_set0(splinter_vlans, 4095);
4050 /* Delete all VLAN devices that we don't need. */
4052 real_devs = vlandev_get_real_devs();
4053 SHASH_FOR_EACH (node, real_devs) {
4054 const struct vlan_real_dev *real_dev = node->data;
4055 const struct vlan_dev *vlan_dev;
4056 bool real_dev_has_splinters;
4058 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4060 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4061 if (!real_dev_has_splinters
4062 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4063 struct netdev *netdev;
4065 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4066 if (!netdev_get_in4(netdev, NULL, NULL) ||
4067 !netdev_get_in6(netdev, NULL)) {
4068 /* It has an IP address configured, so we don't own
4069 * it. Don't delete it. */
4071 vlandev_del(vlan_dev->name);
4073 netdev_close(netdev);
4080 /* Add all VLAN devices that we need. */
4081 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4084 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4085 if (!vlandev_get_name(real_dev_name, vid)) {
4086 vlandev_add(real_dev_name, vid);
4093 sset_destroy(&splinter_ifaces);
4095 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4096 free(splinter_vlans);
4099 return splinter_vlans;
4102 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4105 configure_splinter_port(struct port *port)
4107 struct ofproto *ofproto = port->bridge->ofproto;
4108 ofp_port_t realdev_ofp_port;
4109 const char *realdev_name;
4110 struct iface *vlandev, *realdev;
4112 ofproto_bundle_unregister(port->bridge->ofproto, port);
4114 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4117 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4118 realdev = iface_lookup(port->bridge, realdev_name);
4119 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4121 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4125 static struct ovsrec_port *
4126 synthesize_splinter_port(const char *real_dev_name,
4127 const char *vlan_dev_name, int vid)
4129 struct ovsrec_interface *iface;
4130 struct ovsrec_port *port;
4132 iface = xmalloc(sizeof *iface);
4133 ovsrec_interface_init(iface);
4134 iface->name = xstrdup(vlan_dev_name);
4135 iface->type = "system";
4137 port = xmalloc(sizeof *port);
4138 ovsrec_port_init(port);
4139 port->interfaces = xmemdup(&iface, sizeof iface);
4140 port->n_interfaces = 1;
4141 port->name = xstrdup(vlan_dev_name);
4142 port->vlan_mode = "splinter";
4143 port->tag = xmalloc(sizeof *port->tag);
4146 smap_add(&port->other_config, "realdev", real_dev_name);
4152 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4153 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4154 * 1-bit in the 'splinter_vlans' bitmap. */
4156 add_vlan_splinter_ports(struct bridge *br,
4157 const unsigned long int *splinter_vlans,
4158 struct shash *ports)
4162 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4163 * we're modifying 'ports'. */
4164 for (i = 0; i < br->cfg->n_ports; i++) {
4165 const char *name = br->cfg->ports[i]->name;
4166 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4169 for (j = 0; j < port_cfg->n_interfaces; j++) {
4170 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4172 if (vlan_splinters_is_enabled(iface_cfg)) {
4173 const char *real_dev_name;
4176 real_dev_name = iface_cfg->name;
4177 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4178 const char *vlan_dev_name;
4180 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4182 && !shash_find(ports, vlan_dev_name)) {
4183 shash_add(ports, vlan_dev_name,
4184 synthesize_splinter_port(
4185 real_dev_name, vlan_dev_name, vid));
4194 mirror_refresh_stats(struct mirror *m)
4196 struct ofproto *ofproto = m->bridge->ofproto;
4197 uint64_t tx_packets, tx_bytes;
4200 size_t stat_cnt = 0;
4202 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4203 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4207 if (tx_packets != UINT64_MAX) {
4208 keys[stat_cnt] = "tx_packets";
4209 values[stat_cnt] = tx_packets;
4212 if (tx_bytes != UINT64_MAX) {
4213 keys[stat_cnt] = "tx_bytes";
4214 values[stat_cnt] = tx_bytes;
4218 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);