1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 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_flow_miss_model(const char *opt);
198 static void bridge_configure_datapath_id(struct bridge *);
199 static void bridge_configure_netflow(struct bridge *);
200 static void bridge_configure_forward_bpdu(struct bridge *);
201 static void bridge_configure_mac_table(struct bridge *);
202 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
203 static void bridge_configure_ipfix(struct bridge *);
204 static void bridge_configure_stp(struct bridge *);
205 static void bridge_configure_tables(struct bridge *);
206 static void bridge_configure_dp_desc(struct bridge *);
207 static void bridge_configure_remotes(struct bridge *,
208 const struct sockaddr_in *managers,
210 static void bridge_pick_local_hw_addr(struct bridge *,
211 uint8_t ea[ETH_ADDR_LEN],
212 struct iface **hw_addr_iface);
213 static uint64_t bridge_pick_datapath_id(struct bridge *,
214 const uint8_t bridge_ea[ETH_ADDR_LEN],
215 struct iface *hw_addr_iface);
216 static uint64_t dpid_from_hash(const void *, size_t nbytes);
217 static bool bridge_has_bond_fake_iface(const struct bridge *,
219 static bool port_is_bond_fake_iface(const struct port *);
221 static unixctl_cb_func qos_unixctl_show;
223 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
224 static void port_del_ifaces(struct port *);
225 static void port_destroy(struct port *);
226 static struct port *port_lookup(const struct bridge *, const char *name);
227 static void port_configure(struct port *);
228 static struct lacp_settings *port_configure_lacp(struct port *,
229 struct lacp_settings *);
230 static void port_configure_bond(struct port *, struct bond_settings *);
231 static bool port_is_synthetic(const struct port *);
233 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
234 static void run_system_stats(void);
236 static void bridge_configure_mirrors(struct bridge *);
237 static struct mirror *mirror_create(struct bridge *,
238 const struct ovsrec_mirror *);
239 static void mirror_destroy(struct mirror *);
240 static bool mirror_configure(struct mirror *);
241 static void mirror_refresh_stats(struct mirror *);
243 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
244 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
245 const struct ovsrec_port *);
246 static bool iface_is_internal(const struct ovsrec_interface *iface,
247 const struct ovsrec_bridge *br);
248 static const char *iface_get_type(const struct ovsrec_interface *,
249 const struct ovsrec_bridge *);
250 static void iface_destroy(struct iface *);
251 static struct iface *iface_lookup(const struct bridge *, const char *name);
252 static struct iface *iface_find(const char *name);
253 static struct iface *iface_from_ofp_port(const struct bridge *,
254 ofp_port_t ofp_port);
255 static void iface_set_mac(struct iface *, const uint8_t *);
256 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
257 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
258 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
259 static void iface_configure_cfm(struct iface *);
260 static void iface_refresh_cfm_stats(struct iface *);
261 static void iface_refresh_stats(struct iface *);
262 static void iface_refresh_status(struct iface *);
263 static bool iface_is_synthetic(const struct iface *);
264 static ofp_port_t iface_get_requested_ofp_port(
265 const struct ovsrec_interface *);
266 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
268 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
270 * This is deprecated. It is only for compatibility with broken device drivers
271 * in old versions of Linux that do not properly support VLANs when VLAN
272 * devices are not used. When broken device drivers are no longer in
273 * widespread use, we will delete these interfaces. */
275 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
276 static bool vlan_splinters_enabled_anywhere;
278 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
279 static unsigned long int *collect_splinter_vlans(
280 const struct ovsrec_open_vswitch *);
281 static void configure_splinter_port(struct port *);
282 static void add_vlan_splinter_ports(struct bridge *,
283 const unsigned long int *splinter_vlans,
284 struct shash *ports);
287 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
289 struct shash iface_hints;
290 static bool initialized = false;
297 shash_init(&iface_hints);
300 for (i = 0; i < cfg->n_bridges; i++) {
301 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
304 for (j = 0; j < br_cfg->n_ports; j++) {
305 struct ovsrec_port *port_cfg = br_cfg->ports[j];
308 for (k = 0; k < port_cfg->n_interfaces; k++) {
309 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
310 struct iface_hint *iface_hint;
312 iface_hint = xmalloc(sizeof *iface_hint);
313 iface_hint->br_name = br_cfg->name;
314 iface_hint->br_type = br_cfg->datapath_type;
315 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
317 shash_add(&iface_hints, if_cfg->name, iface_hint);
323 ofproto_init(&iface_hints);
325 shash_destroy_free_data(&iface_hints);
329 /* Public functions. */
331 /* Initializes the bridge module, configuring it to obtain its configuration
332 * from an OVSDB server accessed over 'remote', which should be a string in a
333 * form acceptable to ovsdb_idl_create(). */
335 bridge_init(const char *remote)
337 /* Create connection to database. */
338 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
339 idl_seqno = ovsdb_idl_get_seqno(idl);
340 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
341 ovsdb_idl_verify_write_only(idl);
343 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
345 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
351 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
353 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
355 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
357 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
359 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
378 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
383 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
385 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
390 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
392 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
403 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
405 /* Register unixctl commands. */
406 unixctl_command_register("qos/show", "interface", 1, 1,
407 qos_unixctl_show, NULL);
408 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
409 bridge_unixctl_dump_flows, NULL);
410 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
411 bridge_unixctl_reconnect, NULL);
421 struct bridge *br, *next_br;
423 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
426 ovsdb_idl_destroy(idl);
429 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
430 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
431 * responsible for freeing '*managersp' (with free()).
433 * You may be asking yourself "why does ovs-vswitchd care?", because
434 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
435 * should not be and in fact is not directly involved in that. But
436 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
437 * it has to tell in-band control where the managers are to enable that.
438 * (Thus, only managers connected in-band are collected.)
441 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
442 struct sockaddr_in **managersp, size_t *n_managersp)
444 struct sockaddr_in *managers = NULL;
445 size_t n_managers = 0;
449 /* Collect all of the potential targets from the "targets" columns of the
450 * rows pointed to by "manager_options", excluding any that are
453 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
454 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
456 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
457 sset_find_and_delete(&targets, m->target);
459 sset_add(&targets, m->target);
463 /* Now extract the targets' IP addresses. */
464 if (!sset_is_empty(&targets)) {
467 managers = xmalloc(sset_count(&targets) * sizeof *managers);
468 SSET_FOR_EACH (target, &targets) {
469 struct sockaddr_in *sin = &managers[n_managers];
471 if (stream_parse_target_with_default_port(target,
478 sset_destroy(&targets);
480 *managersp = managers;
481 *n_managersp = n_managers;
485 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
487 unsigned long int *splinter_vlans;
488 struct sockaddr_in *managers;
489 struct bridge *br, *next;
490 int sflow_bridge_number;
493 COVERAGE_INC(bridge_reconfigure);
495 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
496 OFPROTO_FLOW_LIMIT_DEFAULT));
499 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
500 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
502 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
503 "force-miss-model"));
505 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
506 * to 'ovs_cfg', with only very minimal configuration otherwise.
508 * This is mostly an update to bridge data structures. Nothing is pushed
509 * down to ofproto or lower layers. */
510 add_del_bridges(ovs_cfg);
511 splinter_vlans = collect_splinter_vlans(ovs_cfg);
512 HMAP_FOR_EACH (br, node, &all_bridges) {
513 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
514 bridge_del_ports(br, &br->wanted_ports);
516 free(splinter_vlans);
518 /* Start pushing configuration changes down to the ofproto layer:
520 * - Delete ofprotos that are no longer configured.
522 * - Delete ports that are no longer configured.
524 * - Reconfigure existing ports to their desired configurations, or
525 * delete them if not possible.
527 * We have to do all the deletions before we can do any additions, because
528 * the ports to be added might require resources that will be freed up by
529 * deletions (they might especially overlap in name). */
530 bridge_delete_ofprotos();
531 HMAP_FOR_EACH (br, node, &all_bridges) {
533 bridge_delete_or_reconfigure_ports(br);
537 /* Finish pushing configuration changes to the ofproto layer:
539 * - Create ofprotos that are missing.
541 * - Add ports that are missing. */
542 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
546 error = ofproto_create(br->name, br->type, &br->ofproto);
548 VLOG_ERR("failed to create bridge %s: %s", br->name,
549 ovs_strerror(error));
550 shash_destroy(&br->wanted_ports);
555 HMAP_FOR_EACH (br, node, &all_bridges) {
556 bridge_add_ports(br, &br->wanted_ports);
557 shash_destroy(&br->wanted_ports);
560 reconfigure_system_stats(ovs_cfg);
562 /* Complete the configuration. */
563 sflow_bridge_number = 0;
564 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
565 HMAP_FOR_EACH (br, node, &all_bridges) {
568 /* We need the datapath ID early to allow LACP ports to use it as the
569 * default system ID. */
570 bridge_configure_datapath_id(br);
572 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
575 port_configure(port);
577 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
578 iface_configure_cfm(iface);
579 iface_configure_qos(iface, port->cfg->qos);
580 iface_set_mac(iface, port->cfg->fake_bridge ? br->ea : NULL);
581 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
585 bridge_configure_mirrors(br);
586 bridge_configure_forward_bpdu(br);
587 bridge_configure_mac_table(br);
588 bridge_configure_remotes(br, managers, n_managers);
589 bridge_configure_netflow(br);
590 bridge_configure_sflow(br, &sflow_bridge_number);
591 bridge_configure_ipfix(br);
592 bridge_configure_stp(br);
593 bridge_configure_tables(br);
594 bridge_configure_dp_desc(br);
596 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
597 /* XXX: Remove this warning message eventually. */
598 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
599 " moved to the Open_vSwitch table. Ignoring its"
600 " setting in the bridge table.");
605 /* The ofproto-dpif provider does some final reconfiguration in its
606 * ->type_run() function. We have to call it before notifying the database
607 * client that reconfiguration is complete, otherwise there is a very
608 * narrow race window in which e.g. ofproto/trace will not recognize the
609 * new configuration (sometimes this causes unit test failures). */
613 /* Delete ofprotos which aren't configured or have the wrong type. Create
614 * ofprotos which don't exist but need to. */
616 bridge_delete_ofprotos(void)
623 /* Delete ofprotos with no bridge or with the wrong type. */
626 ofproto_enumerate_types(&types);
627 SSET_FOR_EACH (type, &types) {
630 ofproto_enumerate_names(type, &names);
631 SSET_FOR_EACH (name, &names) {
632 br = bridge_lookup(name);
633 if (!br || strcmp(type, br->type)) {
634 ofproto_delete(name, type);
638 sset_destroy(&names);
639 sset_destroy(&types);
643 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
645 if (*n >= *allocated) {
646 ports = x2nrealloc(ports, allocated, sizeof *ports);
648 ports[(*n)++] = port;
653 bridge_delete_or_reconfigure_ports(struct bridge *br)
655 struct ofproto_port ofproto_port;
656 struct ofproto_port_dump dump;
658 /* List of "ofp_port"s to delete. We make a list instead of deleting them
659 * right away because ofproto implementations aren't necessarily able to
660 * iterate through a changing list of ports in an entirely robust way. */
668 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
669 ofp_port_t requested_ofp_port;
672 iface = iface_lookup(br, ofproto_port.name);
674 /* No such iface is configured, so we should delete this
677 * As a corner case exception, keep the port if it's a bond fake
679 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
680 && !strcmp(ofproto_port.type, "internal")) {
686 if (strcmp(ofproto_port.type, iface->type)
687 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
688 /* The interface is the wrong type or can't be configured.
693 /* If the requested OpenFlow port for 'iface' changed, and it's not
694 * already the correct port, then we might want to temporarily delete
695 * this interface, so we can add it back again with the new OpenFlow
697 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
698 if (iface->ofp_port != OFPP_LOCAL &&
699 requested_ofp_port != OFPP_NONE &&
700 requested_ofp_port != iface->ofp_port) {
701 ofp_port_t victim_request;
702 struct iface *victim;
704 /* Check for an existing OpenFlow port currently occupying
705 * 'iface''s requested port number. If there isn't one, then
706 * delete this port. Otherwise we need to consider further. */
707 victim = iface_from_ofp_port(br, requested_ofp_port);
712 /* 'victim' is a port currently using 'iface''s requested port
713 * number. Unless 'victim' specifically requested that port
714 * number, too, then we can delete both 'iface' and 'victim'
715 * temporarily. (We'll add both of them back again later with new
716 * OpenFlow port numbers.)
718 * If 'victim' did request port number 'requested_ofp_port', just
719 * like 'iface', then that's a configuration inconsistency that we
720 * can't resolve. We might as well let it keep its current port
722 victim_request = iface_get_requested_ofp_port(victim->cfg);
723 if (victim_request != requested_ofp_port) {
724 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
725 iface_destroy(victim);
734 iface_destroy(iface);
735 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
738 for (i = 0; i < n; i++) {
739 ofproto_port_del(br->ofproto, del[i]);
745 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
746 bool with_requested_port)
748 struct shash_node *port_node;
750 SHASH_FOR_EACH (port_node, wanted_ports) {
751 const struct ovsrec_port *port_cfg = port_node->data;
754 for (i = 0; i < port_cfg->n_interfaces; i++) {
755 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
756 ofp_port_t requested_ofp_port;
758 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
759 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
760 struct iface *iface = iface_lookup(br, iface_cfg->name);
763 iface_create(br, iface_cfg, port_cfg);
771 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
773 /* First add interfaces that request a particular port number. */
774 bridge_add_ports__(br, wanted_ports, true);
776 /* Then add interfaces that want automatic port number assignment.
777 * We add these afterward to avoid accidentally taking a specifically
778 * requested port number. */
779 bridge_add_ports__(br, wanted_ports, false);
783 port_configure(struct port *port)
785 const struct ovsrec_port *cfg = port->cfg;
786 struct bond_settings bond_settings;
787 struct lacp_settings lacp_settings;
788 struct ofproto_bundle_settings s;
791 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
792 configure_splinter_port(port);
801 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
802 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
803 s.slaves[s.n_slaves++] = iface->ofp_port;
808 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
812 /* Get VLAN trunks. */
815 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
819 if (cfg->vlan_mode) {
820 if (!strcmp(cfg->vlan_mode, "access")) {
821 s.vlan_mode = PORT_VLAN_ACCESS;
822 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
823 s.vlan_mode = PORT_VLAN_TRUNK;
824 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
825 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
826 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
827 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
829 /* This "can't happen" because ovsdb-server should prevent it. */
830 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
831 s.vlan_mode = PORT_VLAN_TRUNK;
835 s.vlan_mode = PORT_VLAN_ACCESS;
837 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
841 s.vlan_mode = PORT_VLAN_TRUNK;
844 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
847 /* Get LACP settings. */
848 s.lacp = port_configure_lacp(port, &lacp_settings);
852 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
853 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
854 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
857 s.lacp_slaves = NULL;
860 /* Get bond settings. */
861 if (s.n_slaves > 1) {
862 s.bond = &bond_settings;
863 port_configure_bond(port, &bond_settings);
866 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
867 netdev_set_miimon_interval(iface->netdev, 0);
872 ofproto_bundle_register(port->bridge->ofproto, port, &s);
881 bridge_configure_flow_miss_model(const char *opt)
883 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
886 if (!strcmp(opt, "with-facets")) {
887 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
888 } else if (!strcmp(opt, "without-facets")) {
889 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
893 ofproto_set_flow_miss_model(model);
896 /* Pick local port hardware address and datapath ID for 'br'. */
898 bridge_configure_datapath_id(struct bridge *br)
900 uint8_t ea[ETH_ADDR_LEN];
902 struct iface *local_iface;
903 struct iface *hw_addr_iface;
906 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
907 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
909 int error = netdev_set_etheraddr(local_iface->netdev, ea);
911 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
912 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
913 "Ethernet address: %s",
914 br->name, ovs_strerror(error));
917 memcpy(br->ea, ea, ETH_ADDR_LEN);
919 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
920 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
921 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
922 ofproto_set_datapath_id(br->ofproto, dpid);
925 dpid_string = xasprintf("%016"PRIx64, dpid);
926 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
930 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
933 bridge_get_allowed_versions(struct bridge *br)
935 if (!br->cfg->n_protocols)
938 return ofputil_versions_from_strings(br->cfg->protocols,
939 br->cfg->n_protocols);
942 /* Set NetFlow configuration on 'br'. */
944 bridge_configure_netflow(struct bridge *br)
946 struct ovsrec_netflow *cfg = br->cfg->netflow;
947 struct netflow_options opts;
950 ofproto_set_netflow(br->ofproto, NULL);
954 memset(&opts, 0, sizeof opts);
956 /* Get default NetFlow configuration from datapath.
957 * Apply overrides from 'cfg'. */
958 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
959 if (cfg->engine_type) {
960 opts.engine_type = *cfg->engine_type;
962 if (cfg->engine_id) {
963 opts.engine_id = *cfg->engine_id;
966 /* Configure active timeout interval. */
967 opts.active_timeout = cfg->active_timeout;
968 if (!opts.active_timeout) {
969 opts.active_timeout = -1;
970 } else if (opts.active_timeout < 0) {
971 VLOG_WARN("bridge %s: active timeout interval set to negative "
972 "value, using default instead (%d seconds)", br->name,
973 NF_ACTIVE_TIMEOUT_DEFAULT);
974 opts.active_timeout = -1;
977 /* Add engine ID to interface number to disambiguate bridgs? */
978 opts.add_id_to_iface = cfg->add_id_to_interface;
979 if (opts.add_id_to_iface) {
980 if (opts.engine_id > 0x7f) {
981 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
982 "another vswitch, choose an engine id less than 128",
985 if (hmap_count(&br->ports) > 508) {
986 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
987 "another port when more than 508 ports are used",
993 sset_init(&opts.collectors);
994 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
997 if (ofproto_set_netflow(br->ofproto, &opts)) {
998 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1000 sset_destroy(&opts.collectors);
1003 /* Set sFlow configuration on 'br'. */
1005 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1007 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1008 struct ovsrec_controller **controllers;
1009 struct ofproto_sflow_options oso;
1010 size_t n_controllers;
1014 ofproto_set_sflow(br->ofproto, NULL);
1018 memset(&oso, 0, sizeof oso);
1020 sset_init(&oso.targets);
1021 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1023 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1024 if (cfg->sampling) {
1025 oso.sampling_rate = *cfg->sampling;
1028 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1030 oso.polling_interval = *cfg->polling;
1033 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1035 oso.header_len = *cfg->header;
1038 oso.sub_id = (*sflow_bridge_number)++;
1039 oso.agent_device = cfg->agent;
1041 oso.control_ip = NULL;
1042 n_controllers = bridge_get_controllers(br, &controllers);
1043 for (i = 0; i < n_controllers; i++) {
1044 if (controllers[i]->local_ip) {
1045 oso.control_ip = controllers[i]->local_ip;
1049 ofproto_set_sflow(br->ofproto, &oso);
1051 sset_destroy(&oso.targets);
1054 /* Returns whether a IPFIX row is valid. */
1056 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1058 return ipfix && ipfix->n_targets > 0;
1061 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1063 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1064 const struct bridge *br)
1066 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1069 /* Set IPFIX configuration on 'br'. */
1071 bridge_configure_ipfix(struct bridge *br)
1073 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1074 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1075 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1076 struct ofproto_ipfix_bridge_exporter_options be_opts;
1077 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1078 size_t n_fe_opts = 0;
1080 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1081 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1086 if (!valid_be_cfg && n_fe_opts == 0) {
1087 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1092 memset(&be_opts, 0, sizeof be_opts);
1094 sset_init(&be_opts.targets);
1095 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1097 if (be_cfg->sampling) {
1098 be_opts.sampling_rate = *be_cfg->sampling;
1100 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1102 if (be_cfg->obs_domain_id) {
1103 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1105 if (be_cfg->obs_point_id) {
1106 be_opts.obs_point_id = *be_cfg->obs_point_id;
1108 if (be_cfg->cache_active_timeout) {
1109 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1111 if (be_cfg->cache_max_flows) {
1112 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1116 if (n_fe_opts > 0) {
1117 struct ofproto_ipfix_flow_exporter_options *opts;
1118 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1120 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1121 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1122 opts->collector_set_id = fe_cfg->id;
1123 sset_init(&opts->targets);
1124 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1125 fe_cfg->ipfix->n_targets);
1126 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1127 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1128 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1129 ? *fe_cfg->ipfix->cache_max_flows : 0;
1135 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1139 sset_destroy(&be_opts.targets);
1142 if (n_fe_opts > 0) {
1143 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1145 for (i = 0; i < n_fe_opts; i++) {
1146 sset_destroy(&opts->targets);
1154 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1155 struct ofproto_port_stp_settings *port_s,
1156 int *port_num_counter, unsigned long *port_num_bitmap)
1158 const char *config_str;
1159 struct iface *iface;
1161 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1162 port_s->enable = false;
1165 port_s->enable = true;
1168 /* STP over bonds is not supported. */
1169 if (!list_is_singleton(&port->ifaces)) {
1170 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1172 port_s->enable = false;
1176 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1178 /* Internal ports shouldn't participate in spanning tree, so
1180 if (!strcmp(iface->type, "internal")) {
1181 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1182 port_s->enable = false;
1186 /* STP on mirror output ports is not supported. */
1187 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1188 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1189 port_s->enable = false;
1193 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1195 unsigned long int port_num = strtoul(config_str, NULL, 0);
1196 int port_idx = port_num - 1;
1198 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1199 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1200 port_s->enable = false;
1204 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1205 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1206 port->name, port_num);
1207 port_s->enable = false;
1210 bitmap_set1(port_num_bitmap, port_idx);
1211 port_s->port_num = port_idx;
1213 if (*port_num_counter >= STP_MAX_PORTS) {
1214 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1215 port_s->enable = false;
1219 port_s->port_num = (*port_num_counter)++;
1222 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1224 port_s->path_cost = strtoul(config_str, NULL, 10);
1226 enum netdev_features current;
1229 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1230 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1231 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1234 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1236 port_s->priority = strtoul(config_str, NULL, 0);
1238 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1242 /* Set spanning tree configuration on 'br'. */
1244 bridge_configure_stp(struct bridge *br)
1246 if (!br->cfg->stp_enable) {
1247 ofproto_set_stp(br->ofproto, NULL);
1249 struct ofproto_stp_settings br_s;
1250 const char *config_str;
1252 int port_num_counter;
1253 unsigned long *port_num_bitmap;
1255 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1257 uint8_t ea[ETH_ADDR_LEN];
1259 if (eth_addr_from_string(config_str, ea)) {
1260 br_s.system_id = eth_addr_to_uint64(ea);
1262 br_s.system_id = eth_addr_to_uint64(br->ea);
1263 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1264 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1267 br_s.system_id = eth_addr_to_uint64(br->ea);
1270 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1272 br_s.priority = strtoul(config_str, NULL, 0);
1274 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1277 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1279 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1281 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1284 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1286 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1288 br_s.max_age = STP_DEFAULT_MAX_AGE;
1291 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1293 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1295 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1298 /* Configure STP on the bridge. */
1299 if (ofproto_set_stp(br->ofproto, &br_s)) {
1300 VLOG_ERR("bridge %s: could not enable STP", br->name);
1304 /* Users must either set the port number with the "stp-port-num"
1305 * configuration on all ports or none. If manual configuration
1306 * is not done, then we allocate them sequentially. */
1307 port_num_counter = 0;
1308 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1309 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1310 struct ofproto_port_stp_settings port_s;
1311 struct iface *iface;
1313 port_configure_stp(br->ofproto, port, &port_s,
1314 &port_num_counter, port_num_bitmap);
1316 /* As bonds are not supported, just apply configuration to
1317 * all interfaces. */
1318 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1319 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1321 VLOG_ERR("port %s: could not enable STP", port->name);
1327 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1328 && port_num_counter) {
1329 VLOG_ERR("bridge %s: must manually configure all STP port "
1330 "IDs or none, disabling", br->name);
1331 ofproto_set_stp(br->ofproto, NULL);
1333 bitmap_free(port_num_bitmap);
1338 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1340 const struct port *port = port_lookup(br, name);
1341 return port && port_is_bond_fake_iface(port);
1345 port_is_bond_fake_iface(const struct port *port)
1347 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1351 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1353 struct bridge *br, *next;
1354 struct shash new_br;
1357 /* Collect new bridges' names and types. */
1358 shash_init(&new_br);
1359 for (i = 0; i < cfg->n_bridges; i++) {
1360 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1361 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1363 if (strchr(br_cfg->name, '/')) {
1364 /* Prevent remote ovsdb-server users from accessing arbitrary
1365 * directories, e.g. consider a bridge named "../../../etc/". */
1366 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1368 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1369 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1373 /* Get rid of deleted bridges or those whose types have changed.
1374 * Update 'cfg' of bridges that still exist. */
1375 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1376 br->cfg = shash_find_data(&new_br, br->name);
1377 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1378 br->cfg->datapath_type))) {
1383 /* Add new bridges. */
1384 for (i = 0; i < cfg->n_bridges; i++) {
1385 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1386 struct bridge *br = bridge_lookup(br_cfg->name);
1388 bridge_create(br_cfg);
1392 shash_destroy(&new_br);
1395 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1396 * Returns 0 if successful, otherwise a positive errno value. */
1398 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1399 struct netdev *netdev)
1401 return netdev_set_config(netdev, &iface_cfg->options);
1404 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1405 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1407 * If successful, returns 0 and stores the network device in '*netdevp'. On
1408 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1410 iface_do_create(const struct bridge *br,
1411 const struct ovsrec_interface *iface_cfg,
1412 const struct ovsrec_port *port_cfg,
1413 ofp_port_t *ofp_portp, struct netdev **netdevp)
1415 struct netdev *netdev = NULL;
1418 if (netdev_is_reserved_name(iface_cfg->name)) {
1419 VLOG_WARN("could not create interface %s, name is reserved",
1425 error = netdev_open(iface_cfg->name,
1426 iface_get_type(iface_cfg, br->cfg), &netdev);
1428 VLOG_WARN("could not open network device %s (%s)",
1429 iface_cfg->name, ovs_strerror(error));
1433 error = iface_set_netdev_config(iface_cfg, netdev);
1438 *ofp_portp = iface_pick_ofport(iface_cfg);
1439 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1444 VLOG_INFO("bridge %s: added interface %s on port %d",
1445 br->name, iface_cfg->name, *ofp_portp);
1447 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1448 || iface_is_internal(iface_cfg, br->cfg)) {
1449 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1457 netdev_close(netdev);
1461 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1462 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1463 * automatically allocated for the iface. Takes ownership of and
1464 * deallocates 'if_cfg'.
1466 * Return true if an iface is successfully created, false otherwise. */
1468 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1469 const struct ovsrec_port *port_cfg)
1471 struct netdev *netdev;
1472 struct iface *iface;
1473 ofp_port_t ofp_port;
1477 /* Do the bits that can fail up front. */
1478 ovs_assert(!iface_lookup(br, iface_cfg->name));
1479 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1481 iface_set_ofport(iface_cfg, OFPP_NONE);
1482 iface_clear_db_record(iface_cfg);
1486 /* Get or create the port structure. */
1487 port = port_lookup(br, port_cfg->name);
1489 port = port_create(br, port_cfg);
1492 /* Create the iface structure. */
1493 iface = xzalloc(sizeof *iface);
1494 list_push_back(&port->ifaces, &iface->port_elem);
1495 hmap_insert(&br->iface_by_name, &iface->name_node,
1496 hash_string(iface_cfg->name, 0));
1498 iface->name = xstrdup(iface_cfg->name);
1499 iface->ofp_port = ofp_port;
1500 iface->netdev = netdev;
1501 iface->type = iface_get_type(iface_cfg, br->cfg);
1502 iface->cfg = iface_cfg;
1503 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1504 hash_ofp_port(ofp_port));
1506 iface_set_ofport(iface->cfg, ofp_port);
1508 /* Populate initial status in database. */
1509 iface_refresh_stats(iface);
1510 iface_refresh_status(iface);
1512 /* Add bond fake iface if necessary. */
1513 if (port_is_bond_fake_iface(port)) {
1514 struct ofproto_port ofproto_port;
1516 if (ofproto_port_query_by_name(br->ofproto, port->name,
1518 struct netdev *netdev;
1521 error = netdev_open(port->name, "internal", &netdev);
1523 ofp_port_t fake_ofp_port = OFPP_NONE;
1524 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1525 netdev_close(netdev);
1527 VLOG_WARN("could not open network device %s (%s)",
1528 port->name, ovs_strerror(error));
1531 /* Already exists, nothing to do. */
1532 ofproto_port_destroy(&ofproto_port);
1539 /* Set forward BPDU option. */
1541 bridge_configure_forward_bpdu(struct bridge *br)
1543 ofproto_set_forward_bpdu(br->ofproto,
1544 smap_get_bool(&br->cfg->other_config,
1549 /* Set MAC learning table configuration for 'br'. */
1551 bridge_configure_mac_table(struct bridge *br)
1553 const char *idle_time_str;
1556 const char *mac_table_size_str;
1559 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1560 idle_time = (idle_time_str && atoi(idle_time_str)
1561 ? atoi(idle_time_str)
1562 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1564 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1565 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1566 ? atoi(mac_table_size_str)
1569 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1573 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1574 struct iface **hw_addr_iface)
1576 struct hmapx mirror_output_ports;
1579 bool found_addr = false;
1583 *hw_addr_iface = NULL;
1585 /* Did the user request a particular MAC? */
1586 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1587 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1588 if (eth_addr_is_multicast(ea)) {
1589 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1590 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1591 } else if (eth_addr_is_zero(ea)) {
1592 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1598 /* Mirror output ports don't participate in picking the local hardware
1599 * address. ofproto can't help us find out whether a given port is a
1600 * mirror output because we haven't configured mirrors yet, so we need to
1601 * accumulate them ourselves. */
1602 hmapx_init(&mirror_output_ports);
1603 for (i = 0; i < br->cfg->n_mirrors; i++) {
1604 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1605 if (m->output_port) {
1606 hmapx_add(&mirror_output_ports, m->output_port);
1610 /* Otherwise choose the minimum non-local MAC address among all of the
1612 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1613 uint8_t iface_ea[ETH_ADDR_LEN];
1614 struct iface *candidate;
1615 struct iface *iface;
1617 /* Mirror output ports don't participate. */
1618 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1622 /* Choose the MAC address to represent the port. */
1624 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1625 /* Find the interface with this Ethernet address (if any) so that
1626 * we can provide the correct devname to the caller. */
1627 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1628 uint8_t candidate_ea[ETH_ADDR_LEN];
1629 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1630 && eth_addr_equals(iface_ea, candidate_ea)) {
1635 /* Choose the interface whose MAC address will represent the port.
1636 * The Linux kernel bonding code always chooses the MAC address of
1637 * the first slave added to a bond, and the Fedora networking
1638 * scripts always add slaves to a bond in alphabetical order, so
1639 * for compatibility we choose the interface with the name that is
1640 * first in alphabetical order. */
1641 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1642 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1647 /* The local port doesn't count (since we're trying to choose its
1648 * MAC address anyway). */
1649 if (iface->ofp_port == OFPP_LOCAL) {
1654 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1660 /* Compare against our current choice. */
1661 if (!eth_addr_is_multicast(iface_ea) &&
1662 !eth_addr_is_local(iface_ea) &&
1663 !eth_addr_is_reserved(iface_ea) &&
1664 !eth_addr_is_zero(iface_ea) &&
1665 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1667 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1668 *hw_addr_iface = iface;
1674 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1675 *hw_addr_iface = NULL;
1678 hmapx_destroy(&mirror_output_ports);
1681 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1682 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1683 * an interface on 'br', then that interface must be passed in as
1684 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1685 * 'hw_addr_iface' must be passed in as a null pointer. */
1687 bridge_pick_datapath_id(struct bridge *br,
1688 const uint8_t bridge_ea[ETH_ADDR_LEN],
1689 struct iface *hw_addr_iface)
1692 * The procedure for choosing a bridge MAC address will, in the most
1693 * ordinary case, also choose a unique MAC that we can use as a datapath
1694 * ID. In some special cases, though, multiple bridges will end up with
1695 * the same MAC address. This is OK for the bridges, but it will confuse
1696 * the OpenFlow controller, because each datapath needs a unique datapath
1699 * Datapath IDs must be unique. It is also very desirable that they be
1700 * stable from one run to the next, so that policy set on a datapath
1703 const char *datapath_id;
1706 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1707 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1711 if (!hw_addr_iface) {
1713 * A purely internal bridge, that is, one that has no non-virtual
1714 * network devices on it at all, is difficult because it has no
1715 * natural unique identifier at all.
1717 * When the host is a XenServer, we handle this case by hashing the
1718 * host's UUID with the name of the bridge. Names of bridges are
1719 * persistent across XenServer reboots, although they can be reused if
1720 * an internal network is destroyed and then a new one is later
1721 * created, so this is fairly effective.
1723 * When the host is not a XenServer, we punt by using a random MAC
1724 * address on each run.
1726 const char *host_uuid = xenserver_get_host_uuid();
1728 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1729 dpid = dpid_from_hash(combined, strlen(combined));
1735 return eth_addr_to_uint64(bridge_ea);
1739 dpid_from_hash(const void *data, size_t n)
1741 uint8_t hash[SHA1_DIGEST_SIZE];
1743 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1744 sha1_bytes(data, n, hash);
1745 eth_addr_mark_random(hash);
1746 return eth_addr_to_uint64(hash);
1750 iface_refresh_status(struct iface *iface)
1754 enum netdev_features current;
1758 uint8_t mac[ETH_ADDR_LEN];
1762 if (iface_is_synthetic(iface)) {
1768 if (!netdev_get_status(iface->netdev, &smap)) {
1769 ovsrec_interface_set_status(iface->cfg, &smap);
1771 ovsrec_interface_set_status(iface->cfg, NULL);
1774 smap_destroy(&smap);
1776 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1777 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1779 ovsrec_interface_set_duplex(iface->cfg,
1780 netdev_features_is_full_duplex(current)
1782 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1784 ovsrec_interface_set_duplex(iface->cfg, NULL);
1785 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1788 error = netdev_get_mtu(iface->netdev, &mtu);
1791 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1793 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1796 error = netdev_get_etheraddr(iface->netdev, mac);
1798 char mac_string[32];
1800 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1801 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1803 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1806 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1807 * if there is no valid ifindex number. */
1808 ifindex64 = netdev_get_ifindex(iface->netdev);
1809 if (ifindex64 < 0) {
1812 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1815 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1818 iface_refresh_cfm_stats(struct iface *iface)
1820 const struct ovsrec_interface *cfg = iface->cfg;
1821 struct ofproto_cfm_status status;
1823 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1824 iface->ofp_port, &status)) {
1825 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1826 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1827 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1828 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1829 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1830 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1832 const char *reasons[CFM_FAULT_N_REASONS];
1833 int64_t cfm_health = status.health;
1834 int64_t cfm_flap_count = status.flap_count;
1835 bool faulted = status.faults != 0;
1838 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1841 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1842 int reason = 1 << i;
1843 if (status.faults & reason) {
1844 reasons[j++] = cfm_fault_reason_to_str(reason);
1847 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1849 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1851 if (status.remote_opstate >= 0) {
1852 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1853 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1855 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1858 ovsrec_interface_set_cfm_remote_mpids(cfg,
1859 (const int64_t *)status.rmps,
1861 if (cfm_health >= 0) {
1862 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1864 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1872 iface_refresh_stats(struct iface *iface)
1874 #define IFACE_STATS \
1875 IFACE_STAT(rx_packets, "rx_packets") \
1876 IFACE_STAT(tx_packets, "tx_packets") \
1877 IFACE_STAT(rx_bytes, "rx_bytes") \
1878 IFACE_STAT(tx_bytes, "tx_bytes") \
1879 IFACE_STAT(rx_dropped, "rx_dropped") \
1880 IFACE_STAT(tx_dropped, "tx_dropped") \
1881 IFACE_STAT(rx_errors, "rx_errors") \
1882 IFACE_STAT(tx_errors, "tx_errors") \
1883 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1884 IFACE_STAT(rx_over_errors, "rx_over_err") \
1885 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1886 IFACE_STAT(collisions, "collisions")
1888 #define IFACE_STAT(MEMBER, NAME) + 1
1889 enum { N_IFACE_STATS = IFACE_STATS };
1891 int64_t values[N_IFACE_STATS];
1892 char *keys[N_IFACE_STATS];
1895 struct netdev_stats stats;
1897 if (iface_is_synthetic(iface)) {
1901 /* Intentionally ignore return value, since errors will set 'stats' to
1902 * all-1s, and we will deal with that correctly below. */
1903 netdev_get_stats(iface->netdev, &stats);
1905 /* Copy statistics into keys[] and values[]. */
1907 #define IFACE_STAT(MEMBER, NAME) \
1908 if (stats.MEMBER != UINT64_MAX) { \
1910 values[n] = stats.MEMBER; \
1915 ovs_assert(n <= N_IFACE_STATS);
1917 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1922 br_refresh_stp_status(struct bridge *br)
1924 struct smap smap = SMAP_INITIALIZER(&smap);
1925 struct ofproto *ofproto = br->ofproto;
1926 struct ofproto_stp_status status;
1928 if (ofproto_get_stp_status(ofproto, &status)) {
1932 if (!status.enabled) {
1933 ovsrec_bridge_set_status(br->cfg, NULL);
1937 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1938 STP_ID_ARGS(status.bridge_id));
1939 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1940 STP_ID_ARGS(status.designated_root));
1941 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1943 ovsrec_bridge_set_status(br->cfg, &smap);
1944 smap_destroy(&smap);
1948 port_refresh_stp_status(struct port *port)
1950 struct ofproto *ofproto = port->bridge->ofproto;
1951 struct iface *iface;
1952 struct ofproto_port_stp_status status;
1955 if (port_is_synthetic(port)) {
1959 /* STP doesn't currently support bonds. */
1960 if (!list_is_singleton(&port->ifaces)) {
1961 ovsrec_port_set_status(port->cfg, NULL);
1965 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1966 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1970 if (!status.enabled) {
1971 ovsrec_port_set_status(port->cfg, NULL);
1975 /* Set Status column. */
1977 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1978 smap_add(&smap, "stp_state", stp_state_name(status.state));
1979 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1980 smap_add(&smap, "stp_role", stp_role_name(status.role));
1981 ovsrec_port_set_status(port->cfg, &smap);
1982 smap_destroy(&smap);
1986 port_refresh_stp_stats(struct port *port)
1988 struct ofproto *ofproto = port->bridge->ofproto;
1989 struct iface *iface;
1990 struct ofproto_port_stp_stats stats;
1992 int64_t int_values[3];
1994 if (port_is_synthetic(port)) {
1998 /* STP doesn't currently support bonds. */
1999 if (!list_is_singleton(&port->ifaces)) {
2003 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2004 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2008 if (!stats.enabled) {
2009 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2013 /* Set Statistics column. */
2014 keys[0] = "stp_tx_count";
2015 int_values[0] = stats.tx_count;
2016 keys[1] = "stp_rx_count";
2017 int_values[1] = stats.rx_count;
2018 keys[2] = "stp_error_count";
2019 int_values[2] = stats.error_count;
2021 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2022 ARRAY_SIZE(int_values));
2026 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2028 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2032 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2034 bool enable = enable_system_stats(cfg);
2036 system_stats_enable(enable);
2038 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2043 run_system_stats(void)
2045 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2048 stats = system_stats_run();
2050 struct ovsdb_idl_txn *txn;
2051 struct ovsdb_datum datum;
2053 txn = ovsdb_idl_txn_create(idl);
2054 ovsdb_datum_from_smap(&datum, stats);
2055 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2057 ovsdb_idl_txn_commit(txn);
2058 ovsdb_idl_txn_destroy(txn);
2065 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2068 case OFPCR12_ROLE_EQUAL:
2070 case OFPCR12_ROLE_MASTER:
2072 case OFPCR12_ROLE_SLAVE:
2074 case OFPCR12_ROLE_NOCHANGE:
2076 return "*** INVALID ROLE ***";
2081 refresh_controller_status(void)
2085 const struct ovsrec_controller *cfg;
2089 /* Accumulate status for controllers on all bridges. */
2090 HMAP_FOR_EACH (br, node, &all_bridges) {
2091 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2094 /* Update each controller in the database with current status. */
2095 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2096 struct ofproto_controller_info *cinfo =
2097 shash_find_data(&info, cfg->target);
2100 struct smap smap = SMAP_INITIALIZER(&smap);
2101 const char **values = cinfo->pairs.values;
2102 const char **keys = cinfo->pairs.keys;
2105 for (i = 0; i < cinfo->pairs.n; i++) {
2106 smap_add(&smap, keys[i], values[i]);
2109 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2110 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2112 ovsrec_controller_set_status(cfg, &smap);
2113 smap_destroy(&smap);
2115 ovsrec_controller_set_is_connected(cfg, false);
2116 ovsrec_controller_set_role(cfg, NULL);
2117 ovsrec_controller_set_status(cfg, NULL);
2121 ofproto_free_ofproto_controller_info(&info);
2126 * Some information in the database must be kept as up-to-date as possible to
2127 * allow controllers to respond rapidly to network outages. We call these
2128 * statistics "instant" stats.
2130 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2131 * assuming that they've changed. The only means we have to determine whether
2132 * they have changed are:
2134 * - Try to commit changes to the database. If nothing changed, then
2135 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2138 * - instant_stats_run() is called late in the run loop, after anything that
2139 * might change any of the instant stats.
2141 * We use these two facts together to avoid waking the process up every
2142 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2145 /* Minimum interval between writing updates to the instant stats to the
2147 #define INSTANT_INTERVAL_MSEC 100
2149 /* Current instant stats database transaction, NULL if there is no ongoing
2151 static struct ovsdb_idl_txn *instant_txn;
2153 /* Next time (in msec on monotonic clock) at which we will update the instant
2155 static long long int instant_next_txn = LLONG_MIN;
2157 /* True if the run loop has run since we last saw that the instant stats were
2158 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2159 * to refresh the instant stats. */
2160 static bool instant_stats_could_have_changed;
2163 instant_stats_run(void)
2165 enum ovsdb_idl_txn_status status;
2167 instant_stats_could_have_changed = true;
2173 if (time_msec() < instant_next_txn) {
2176 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2178 seq = seq_read(connectivity_seq_get());
2179 if (seq == connectivity_seqno) {
2182 connectivity_seqno = seq;
2184 instant_txn = ovsdb_idl_txn_create(idl);
2185 HMAP_FOR_EACH (br, node, &all_bridges) {
2186 struct iface *iface;
2189 br_refresh_stp_status(br);
2191 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2192 port_refresh_stp_status(port);
2195 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2196 enum netdev_flags flags;
2198 const char *link_state;
2199 int64_t link_resets;
2202 if (iface_is_synthetic(iface)) {
2206 current = ofproto_port_is_lacp_current(br->ofproto,
2210 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2212 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2215 error = netdev_get_flags(iface->netdev, &flags);
2217 const char *state = flags & NETDEV_UP ? "up" : "down";
2218 ovsrec_interface_set_admin_state(iface->cfg, state);
2220 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2223 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2224 ovsrec_interface_set_link_state(iface->cfg, link_state);
2226 link_resets = netdev_get_carrier_resets(iface->netdev);
2227 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2229 iface_refresh_cfm_stats(iface);
2232 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2234 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2235 smap_destroy(&smap);
2240 status = ovsdb_idl_txn_commit(instant_txn);
2241 if (status != TXN_INCOMPLETE) {
2242 ovsdb_idl_txn_destroy(instant_txn);
2245 if (status == TXN_UNCHANGED) {
2246 instant_stats_could_have_changed = false;
2251 instant_stats_wait(void)
2254 ovsdb_idl_txn_wait(instant_txn);
2255 } else if (instant_stats_could_have_changed) {
2256 poll_timer_wait_until(instant_next_txn);
2267 /* Let each datapath type do the work that it needs to do. */
2269 ofproto_enumerate_types(&types);
2270 SSET_FOR_EACH (type, &types) {
2271 ofproto_type_run(type);
2273 sset_destroy(&types);
2275 /* Let each bridge do the work that it needs to do. */
2276 HMAP_FOR_EACH (br, node, &all_bridges) {
2277 ofproto_run(br->ofproto);
2284 static struct ovsrec_open_vswitch null_cfg;
2285 const struct ovsrec_open_vswitch *cfg;
2287 bool vlan_splinters_changed;
2290 ovsrec_open_vswitch_init(&null_cfg);
2294 if (ovsdb_idl_is_lock_contended(idl)) {
2295 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2296 struct bridge *br, *next_br;
2298 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2299 "disabling this process (pid %ld) until it goes away",
2300 (long int) getpid());
2302 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2305 /* Since we will not be running system_stats_run() in this process
2306 * with the current situation of multiple ovs-vswitchd daemons,
2307 * disable system stats collection. */
2308 system_stats_enable(false);
2310 } else if (!ovsdb_idl_has_lock(idl)) {
2313 cfg = ovsrec_open_vswitch_first(idl);
2315 /* Initialize the ofproto library. This only needs to run once, but
2316 * it must be done after the configuration is set. If the
2317 * initialization has already occurred, bridge_init_ofproto()
2318 * returns immediately. */
2319 bridge_init_ofproto(cfg);
2321 /* Once the value of flow-restore-wait is false, we no longer should
2322 * check its value from the database. */
2323 if (cfg && ofproto_get_flow_restore_wait()) {
2324 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2325 "flow-restore-wait", false));
2330 /* Re-configure SSL. We do this on every trip through the main loop,
2331 * instead of just when the database changes, because the contents of the
2332 * key and certificate files can change without the database changing.
2334 * We do this before bridge_reconfigure() because that function might
2335 * initiate SSL connections and thus requires SSL to be configured. */
2336 if (cfg && cfg->ssl) {
2337 const struct ovsrec_ssl *ssl = cfg->ssl;
2339 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2340 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2343 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2344 * usage has changed. */
2345 vlan_splinters_changed = false;
2346 if (vlan_splinters_enabled_anywhere) {
2347 HMAP_FOR_EACH (br, node, &all_bridges) {
2348 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2349 vlan_splinters_changed = true;
2355 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2356 struct ovsdb_idl_txn *txn;
2358 idl_seqno = ovsdb_idl_get_seqno(idl);
2359 txn = ovsdb_idl_txn_create(idl);
2360 bridge_reconfigure(cfg ? cfg : &null_cfg);
2363 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2366 /* If we are completing our initial configuration for this run
2367 * of ovs-vswitchd, then keep the transaction around to monitor
2368 * it for completion. */
2369 if (initial_config_done) {
2370 ovsdb_idl_txn_commit(txn);
2371 ovsdb_idl_txn_destroy(txn);
2373 initial_config_done = true;
2374 daemonize_txn = txn;
2378 if (daemonize_txn) {
2379 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2380 if (status != TXN_INCOMPLETE) {
2381 ovsdb_idl_txn_destroy(daemonize_txn);
2382 daemonize_txn = NULL;
2384 /* ovs-vswitchd has completed initialization, so allow the
2385 * process that forked us to exit successfully. */
2386 daemonize_complete();
2388 vlog_enable_async();
2390 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2394 /* Refresh interface and mirror stats if necessary. */
2395 if (time_msec() >= iface_stats_timer) {
2397 struct ovsdb_idl_txn *txn;
2399 txn = ovsdb_idl_txn_create(idl);
2400 HMAP_FOR_EACH (br, node, &all_bridges) {
2404 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2405 struct iface *iface;
2407 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2408 iface_refresh_stats(iface);
2409 iface_refresh_status(iface);
2412 port_refresh_stp_stats(port);
2415 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2416 mirror_refresh_stats(m);
2420 refresh_controller_status();
2421 ovsdb_idl_txn_commit(txn);
2422 ovsdb_idl_txn_destroy(txn); /* XXX */
2425 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2429 instant_stats_run();
2438 ovsdb_idl_wait(idl);
2439 if (daemonize_txn) {
2440 ovsdb_idl_txn_wait(daemonize_txn);
2444 ofproto_enumerate_types(&types);
2445 SSET_FOR_EACH (type, &types) {
2446 ofproto_type_wait(type);
2448 sset_destroy(&types);
2450 if (!hmap_is_empty(&all_bridges)) {
2453 HMAP_FOR_EACH (br, node, &all_bridges) {
2454 ofproto_wait(br->ofproto);
2456 poll_timer_wait_until(iface_stats_timer);
2459 system_stats_wait();
2460 instant_stats_wait();
2463 /* Adds some memory usage statistics for bridges into 'usage', for use with
2464 * memory_report(). */
2466 bridge_get_memory_usage(struct simap *usage)
2473 ofproto_enumerate_types(&types);
2474 SSET_FOR_EACH (type, &types) {
2475 ofproto_type_get_memory_usage(type, usage);
2477 sset_destroy(&types);
2479 HMAP_FOR_EACH (br, node, &all_bridges) {
2480 ofproto_get_memory_usage(br->ofproto, usage);
2484 /* QoS unixctl user interface functions. */
2486 struct qos_unixctl_show_cbdata {
2488 struct iface *iface;
2492 qos_unixctl_show_queue(unsigned int queue_id,
2493 const struct smap *details,
2494 struct iface *iface,
2497 struct netdev_queue_stats stats;
2498 struct smap_node *node;
2501 ds_put_cstr(ds, "\n");
2503 ds_put_format(ds, "Queue %u:\n", queue_id);
2505 ds_put_cstr(ds, "Default:\n");
2508 SMAP_FOR_EACH (node, details) {
2509 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2512 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2514 if (stats.tx_packets != UINT64_MAX) {
2515 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2518 if (stats.tx_bytes != UINT64_MAX) {
2519 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2522 if (stats.tx_errors != UINT64_MAX) {
2523 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2526 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2527 queue_id, ovs_strerror(error));
2532 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2533 const char *argv[], void *aux OVS_UNUSED)
2535 struct ds ds = DS_EMPTY_INITIALIZER;
2536 struct smap smap = SMAP_INITIALIZER(&smap);
2537 struct iface *iface;
2539 struct smap_node *node;
2541 iface = iface_find(argv[1]);
2543 unixctl_command_reply_error(conn, "no such interface");
2547 netdev_get_qos(iface->netdev, &type, &smap);
2549 if (*type != '\0') {
2550 struct netdev_queue_dump dump;
2551 struct smap details;
2552 unsigned int queue_id;
2554 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2556 SMAP_FOR_EACH (node, &smap) {
2557 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2560 smap_init(&details);
2561 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2562 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2564 smap_destroy(&details);
2566 unixctl_command_reply(conn, ds_cstr(&ds));
2568 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2569 unixctl_command_reply_error(conn, ds_cstr(&ds));
2572 smap_destroy(&smap);
2576 /* Bridge reconfiguration functions. */
2578 bridge_create(const struct ovsrec_bridge *br_cfg)
2582 ovs_assert(!bridge_lookup(br_cfg->name));
2583 br = xzalloc(sizeof *br);
2585 br->name = xstrdup(br_cfg->name);
2586 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2589 /* Derive the default Ethernet address from the bridge's UUID. This should
2590 * be unique and it will be stable between ovs-vswitchd runs. */
2591 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2592 eth_addr_mark_random(br->default_ea);
2594 hmap_init(&br->ports);
2595 hmap_init(&br->ifaces);
2596 hmap_init(&br->iface_by_name);
2597 hmap_init(&br->mirrors);
2599 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2603 bridge_destroy(struct bridge *br)
2606 struct mirror *mirror, *next_mirror;
2607 struct port *port, *next_port;
2609 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2612 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2613 mirror_destroy(mirror);
2616 hmap_remove(&all_bridges, &br->node);
2617 ofproto_destroy(br->ofproto);
2618 hmap_destroy(&br->ifaces);
2619 hmap_destroy(&br->ports);
2620 hmap_destroy(&br->iface_by_name);
2621 hmap_destroy(&br->mirrors);
2628 static struct bridge *
2629 bridge_lookup(const char *name)
2633 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2634 if (!strcmp(br->name, name)) {
2641 /* Handle requests for a listing of all flows known by the OpenFlow
2642 * stack, including those normally hidden. */
2644 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2645 const char *argv[], void *aux OVS_UNUSED)
2650 br = bridge_lookup(argv[1]);
2652 unixctl_command_reply_error(conn, "Unknown bridge");
2657 ofproto_get_all_flows(br->ofproto, &results);
2659 unixctl_command_reply(conn, ds_cstr(&results));
2660 ds_destroy(&results);
2663 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2664 * connections and reconnect. If BRIDGE is not specified, then all bridges
2665 * drop their controller connections and reconnect. */
2667 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2668 const char *argv[], void *aux OVS_UNUSED)
2672 br = bridge_lookup(argv[1]);
2674 unixctl_command_reply_error(conn, "Unknown bridge");
2677 ofproto_reconnect_controllers(br->ofproto);
2679 HMAP_FOR_EACH (br, node, &all_bridges) {
2680 ofproto_reconnect_controllers(br->ofproto);
2683 unixctl_command_reply(conn, NULL);
2687 bridge_get_controllers(const struct bridge *br,
2688 struct ovsrec_controller ***controllersp)
2690 struct ovsrec_controller **controllers;
2691 size_t n_controllers;
2693 controllers = br->cfg->controller;
2694 n_controllers = br->cfg->n_controller;
2696 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2702 *controllersp = controllers;
2704 return n_controllers;
2708 bridge_collect_wanted_ports(struct bridge *br,
2709 const unsigned long int *splinter_vlans,
2710 struct shash *wanted_ports)
2714 shash_init(wanted_ports);
2716 for (i = 0; i < br->cfg->n_ports; i++) {
2717 const char *name = br->cfg->ports[i]->name;
2718 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2719 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2723 if (bridge_get_controllers(br, NULL)
2724 && !shash_find(wanted_ports, br->name)) {
2725 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2726 br->name, br->name);
2728 ovsrec_interface_init(&br->synth_local_iface);
2729 ovsrec_port_init(&br->synth_local_port);
2731 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2732 br->synth_local_port.n_interfaces = 1;
2733 br->synth_local_port.name = br->name;
2735 br->synth_local_iface.name = br->name;
2736 br->synth_local_iface.type = "internal";
2738 br->synth_local_ifacep = &br->synth_local_iface;
2740 shash_add(wanted_ports, br->name, &br->synth_local_port);
2743 if (splinter_vlans) {
2744 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2748 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2749 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2750 * 'br' needs to complete its configuration. */
2752 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2754 struct shash_node *port_node;
2755 struct port *port, *next;
2757 /* Get rid of deleted ports.
2758 * Get rid of deleted interfaces on ports that still exist. */
2759 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2760 port->cfg = shash_find_data(wanted_ports, port->name);
2764 port_del_ifaces(port);
2768 /* Update iface->cfg and iface->type in interfaces that still exist. */
2769 SHASH_FOR_EACH (port_node, wanted_ports) {
2770 const struct ovsrec_port *port = port_node->data;
2773 for (i = 0; i < port->n_interfaces; i++) {
2774 const struct ovsrec_interface *cfg = port->interfaces[i];
2775 struct iface *iface = iface_lookup(br, cfg->name);
2776 const char *type = iface_get_type(cfg, br->cfg);
2781 } else if (!strcmp(type, "null")) {
2782 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2783 " may be removed in February 2013. Please email"
2784 " dev@openvswitch.org with concerns.",
2787 /* We will add new interfaces later. */
2793 /* Initializes 'oc' appropriately as a management service controller for
2796 * The caller must free oc->target when it is no longer needed. */
2798 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2799 struct ofproto_controller *oc)
2801 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2802 oc->max_backoff = 0;
2803 oc->probe_interval = 60;
2804 oc->band = OFPROTO_OUT_OF_BAND;
2806 oc->burst_limit = 0;
2807 oc->enable_async_msgs = true;
2810 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2812 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2813 struct ofproto_controller *oc)
2817 oc->target = c->target;
2818 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2819 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2820 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2821 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2822 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2823 oc->burst_limit = (c->controller_burst_limit
2824 ? *c->controller_burst_limit : 0);
2825 oc->enable_async_msgs = (!c->enable_async_messages
2826 || *c->enable_async_messages);
2827 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2828 if (dscp < 0 || dscp > 63) {
2829 dscp = DSCP_DEFAULT;
2834 /* Configures the IP stack for 'br''s local interface properly according to the
2835 * configuration in 'c'. */
2837 bridge_configure_local_iface_netdev(struct bridge *br,
2838 struct ovsrec_controller *c)
2840 struct netdev *netdev;
2841 struct in_addr mask, gateway;
2843 struct iface *local_iface;
2846 /* If there's no local interface or no IP address, give up. */
2847 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2848 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2852 /* Bring up the local interface. */
2853 netdev = local_iface->netdev;
2854 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2856 /* Configure the IP address and netmask. */
2857 if (!c->local_netmask
2858 || !inet_aton(c->local_netmask, &mask)
2860 mask.s_addr = guess_netmask(ip.s_addr);
2862 if (!netdev_set_in4(netdev, ip, mask)) {
2863 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2864 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2867 /* Configure the default gateway. */
2868 if (c->local_gateway
2869 && inet_aton(c->local_gateway, &gateway)
2870 && gateway.s_addr) {
2871 if (!netdev_add_router(netdev, gateway)) {
2872 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2873 br->name, IP_ARGS(gateway.s_addr));
2878 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2879 * in either string are treated as equal to any number of slashes in the other,
2880 * e.g. "x///y" is equal to "x/y".
2882 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2883 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2884 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2885 * 'b' against a prefix of 'a'.
2888 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2890 const char *b_start = b;
2892 if (b - b_start >= b_stoplen) {
2894 } else if (*a != *b) {
2896 } else if (*a == '/') {
2897 a += strspn(a, "/");
2898 b += strspn(b, "/");
2899 } else if (*a == '\0') {
2909 bridge_configure_remotes(struct bridge *br,
2910 const struct sockaddr_in *managers, size_t n_managers)
2912 bool disable_in_band;
2914 struct ovsrec_controller **controllers;
2915 size_t n_controllers;
2917 enum ofproto_fail_mode fail_mode;
2919 struct ofproto_controller *ocs;
2923 /* Check if we should disable in-band control on this bridge. */
2924 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2927 /* Set OpenFlow queue ID for in-band control. */
2928 ofproto_set_in_band_queue(br->ofproto,
2929 smap_get_int(&br->cfg->other_config,
2930 "in-band-queue", -1));
2932 if (disable_in_band) {
2933 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2935 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2938 n_controllers = bridge_get_controllers(br, &controllers);
2940 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2943 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2944 for (i = 0; i < n_controllers; i++) {
2945 struct ovsrec_controller *c = controllers[i];
2947 if (!strncmp(c->target, "punix:", 6)
2948 || !strncmp(c->target, "unix:", 5)) {
2949 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2952 if (!strncmp(c->target, "unix:", 5)) {
2953 /* Connect to a listening socket */
2954 whitelist = xasprintf("unix:%s/", ovs_rundir());
2955 if (strchr(c->target, '/') &&
2956 !equal_pathnames(c->target, whitelist,
2957 strlen(whitelist))) {
2958 /* Absolute path specified, but not in ovs_rundir */
2959 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2960 "controller \"%s\" due to possibility for "
2961 "remote exploit. Instead, specify socket "
2962 "in whitelisted \"%s\" or connect to "
2963 "\"unix:%s/%s.mgmt\" (which is always "
2964 "available without special configuration).",
2965 br->name, c->target, whitelist,
2966 ovs_rundir(), br->name);
2971 whitelist = xasprintf("punix:%s/%s.controller",
2972 ovs_rundir(), br->name);
2973 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2974 /* Prevent remote ovsdb-server users from accessing
2975 * arbitrary Unix domain sockets and overwriting arbitrary
2977 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2978 "controller \"%s\" due to possibility of "
2979 "overwriting local files. Instead, specify "
2980 "whitelisted \"%s\" or connect to "
2981 "\"unix:%s/%s.mgmt\" (which is always "
2982 "available without special configuration).",
2983 br->name, c->target, whitelist,
2984 ovs_rundir(), br->name);
2993 bridge_configure_local_iface_netdev(br, c);
2994 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2995 if (disable_in_band) {
2996 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3001 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3002 bridge_get_allowed_versions(br));
3003 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3006 /* Set the fail-mode. */
3007 fail_mode = !br->cfg->fail_mode
3008 || !strcmp(br->cfg->fail_mode, "standalone")
3009 ? OFPROTO_FAIL_STANDALONE
3010 : OFPROTO_FAIL_SECURE;
3011 ofproto_set_fail_mode(br->ofproto, fail_mode);
3013 /* Configure OpenFlow controller connection snooping. */
3014 if (!ofproto_has_snoops(br->ofproto)) {
3018 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3019 ovs_rundir(), br->name));
3020 ofproto_set_snoops(br->ofproto, &snoops);
3021 sset_destroy(&snoops);
3026 bridge_configure_tables(struct bridge *br)
3028 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3032 n_tables = ofproto_get_n_tables(br->ofproto);
3034 for (i = 0; i < n_tables; i++) {
3035 struct ofproto_table_settings s;
3038 s.max_flows = UINT_MAX;
3041 s.n_prefix_fields = 0;
3042 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3044 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3045 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3048 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3049 s.max_flows = *cfg->flow_limit;
3051 if (cfg->overflow_policy
3052 && !strcmp(cfg->overflow_policy, "evict")) {
3054 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3055 for (k = 0; k < cfg->n_groups; k++) {
3056 const char *string = cfg->groups[k];
3059 msg = mf_parse_subfield__(&s.groups[k], &string);
3061 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3062 "'groups' (%s)", br->name, i, msg);
3064 } else if (*string) {
3065 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3066 "element '%s' contains trailing garbage",
3067 br->name, i, cfg->groups[k]);
3073 /* Prefix lookup fields. */
3074 s.n_prefix_fields = 0;
3075 for (k = 0; k < cfg->n_prefixes; k++) {
3076 const char *name = cfg->prefixes[k];
3077 const struct mf_field *mf = mf_from_name(name);
3079 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3083 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3084 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3085 "%s", br->name, name);
3088 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3089 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3090 "field not used: %s", br->name, name);
3093 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3095 if (s.n_prefix_fields > 0) {
3097 struct ds ds = DS_EMPTY_INITIALIZER;
3098 for (k = 0; k < s.n_prefix_fields; k++) {
3100 ds_put_char(&ds, ',');
3102 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3104 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3105 br->name, i, ds_cstr(&ds));
3110 ofproto_configure_table(br->ofproto, i, &s);
3114 for (; j < br->cfg->n_flow_tables; j++) {
3115 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3116 "%"PRId64" not supported by this datapath", br->name,
3117 br->cfg->key_flow_tables[j]);
3122 bridge_configure_dp_desc(struct bridge *br)
3124 ofproto_set_dp_desc(br->ofproto,
3125 smap_get(&br->cfg->other_config, "dp-desc"));
3128 /* Port functions. */
3130 static void iface_destroy__(struct iface *);
3132 static struct port *
3133 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3137 port = xzalloc(sizeof *port);
3139 port->name = xstrdup(cfg->name);
3141 list_init(&port->ifaces);
3143 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3147 /* Deletes interfaces from 'port' that are no longer configured for it. */
3149 port_del_ifaces(struct port *port)
3151 struct iface *iface, *next;
3152 struct sset new_ifaces;
3155 /* Collect list of new interfaces. */
3156 sset_init(&new_ifaces);
3157 for (i = 0; i < port->cfg->n_interfaces; i++) {
3158 const char *name = port->cfg->interfaces[i]->name;
3159 const char *type = port->cfg->interfaces[i]->type;
3160 if (strcmp(type, "null")) {
3161 sset_add(&new_ifaces, name);
3165 /* Get rid of deleted interfaces. */
3166 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3167 if (!sset_contains(&new_ifaces, iface->name)) {
3168 iface_destroy(iface);
3172 sset_destroy(&new_ifaces);
3176 port_destroy(struct port *port)
3179 struct bridge *br = port->bridge;
3180 struct iface *iface, *next;
3183 ofproto_bundle_unregister(br->ofproto, port);
3186 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3187 iface_destroy__(iface);
3190 hmap_remove(&br->ports, &port->hmap_node);
3196 static struct port *
3197 port_lookup(const struct bridge *br, const char *name)
3201 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3203 if (!strcmp(port->name, name)) {
3211 enable_lacp(struct port *port, bool *activep)
3213 if (!port->cfg->lacp) {
3214 /* XXX when LACP implementation has been sufficiently tested, enable by
3215 * default and make active on bonded ports. */
3217 } else if (!strcmp(port->cfg->lacp, "off")) {
3219 } else if (!strcmp(port->cfg->lacp, "active")) {
3222 } else if (!strcmp(port->cfg->lacp, "passive")) {
3226 VLOG_WARN("port %s: unknown LACP mode %s",
3227 port->name, port->cfg->lacp);
3232 static struct lacp_settings *
3233 port_configure_lacp(struct port *port, struct lacp_settings *s)
3235 const char *lacp_time, *system_id;
3238 if (!enable_lacp(port, &s->active)) {
3242 s->name = port->name;
3244 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3246 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3247 ETH_ADDR_SCAN_ARGS(s->id))) {
3248 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3249 " address.", port->name, system_id);
3253 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3256 if (eth_addr_is_zero(s->id)) {
3257 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3261 /* Prefer bondable links if unspecified. */
3262 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3264 s->priority = (priority > 0 && priority <= UINT16_MAX
3266 : UINT16_MAX - !list_is_short(&port->ifaces));
3268 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3269 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3271 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3272 "lacp-fallback-ab", false);
3278 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3280 int priority, portid, key;
3282 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3283 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3285 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3287 if (portid <= 0 || portid > UINT16_MAX) {
3288 portid = ofp_to_u16(iface->ofp_port);
3291 if (priority <= 0 || priority > UINT16_MAX) {
3292 priority = UINT16_MAX;
3295 if (key < 0 || key > UINT16_MAX) {
3299 s->name = iface->name;
3301 s->priority = priority;
3306 port_configure_bond(struct port *port, struct bond_settings *s)
3308 const char *detect_s;
3309 struct iface *iface;
3310 int miimon_interval;
3312 s->name = port->name;
3314 if (port->cfg->bond_mode) {
3315 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3316 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3317 port->name, port->cfg->bond_mode,
3318 bond_mode_to_string(s->balance));
3321 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3323 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3324 * active-backup. At some point we should remove this warning. */
3325 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3326 " in previous versions, the default bond_mode was"
3327 " balance-slb", port->name,
3328 bond_mode_to_string(s->balance));
3330 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3331 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3332 "please use another bond type or disable flood_vlans",
3336 miimon_interval = smap_get_int(&port->cfg->other_config,
3337 "bond-miimon-interval", 0);
3338 if (miimon_interval <= 0) {
3339 miimon_interval = 200;
3342 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3343 if (!detect_s || !strcmp(detect_s, "carrier")) {
3344 miimon_interval = 0;
3345 } else if (strcmp(detect_s, "miimon")) {
3346 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3347 "defaulting to carrier", port->name, detect_s);
3348 miimon_interval = 0;
3351 s->up_delay = MAX(0, port->cfg->bond_updelay);
3352 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3353 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3354 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3355 "bond-rebalance-interval", 10000);
3356 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3357 s->rebalance_interval = 1000;
3360 s->fake_iface = port->cfg->bond_fake_iface;
3362 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3363 "lacp-fallback-ab", false);
3365 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3366 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3370 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3371 * instead of obtaining it from the database. */
3373 port_is_synthetic(const struct port *port)
3375 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3378 /* Interface functions. */
3381 iface_is_internal(const struct ovsrec_interface *iface,
3382 const struct ovsrec_bridge *br)
3384 /* The local port and "internal" ports are always "internal". */
3385 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3388 /* Returns the correct network device type for interface 'iface' in bridge
3391 iface_get_type(const struct ovsrec_interface *iface,
3392 const struct ovsrec_bridge *br)
3396 /* The local port always has type "internal". Other ports take
3397 * their type from the database and default to "system" if none is
3399 if (iface_is_internal(iface, br)) {
3402 type = iface->type[0] ? iface->type : "system";
3405 return ofproto_port_open_type(br->datapath_type, type);
3409 iface_destroy__(struct iface *iface)
3412 struct port *port = iface->port;
3413 struct bridge *br = port->bridge;
3415 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3416 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3419 if (iface->ofp_port != OFPP_NONE) {
3420 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3423 list_remove(&iface->port_elem);
3424 hmap_remove(&br->iface_by_name, &iface->name_node);
3426 netdev_close(iface->netdev);
3434 iface_destroy(struct iface *iface)
3437 struct port *port = iface->port;
3439 iface_destroy__(iface);
3440 if (list_is_empty(&port->ifaces)) {
3446 static struct iface *
3447 iface_lookup(const struct bridge *br, const char *name)
3449 struct iface *iface;
3451 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3452 &br->iface_by_name) {
3453 if (!strcmp(iface->name, name)) {
3461 static struct iface *
3462 iface_find(const char *name)
3464 const struct bridge *br;
3466 HMAP_FOR_EACH (br, node, &all_bridges) {
3467 struct iface *iface = iface_lookup(br, name);
3476 static struct iface *
3477 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3479 struct iface *iface;
3481 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3483 if (iface->ofp_port == ofp_port) {
3490 /* Set Ethernet address of 'iface', if one is specified in the configuration
3493 iface_set_mac(struct iface *iface, const uint8_t *mac)
3495 uint8_t ea[ETH_ADDR_LEN];
3497 if (strcmp(iface->type, "internal")) {
3501 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3506 if (iface->ofp_port == OFPP_LOCAL) {
3507 VLOG_ERR("interface %s: ignoring mac in Interface record "
3508 "(use Bridge record to set local port's mac)",
3510 } else if (eth_addr_is_multicast(mac)) {
3511 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3514 int error = netdev_set_etheraddr(iface->netdev, mac);
3516 VLOG_ERR("interface %s: setting MAC failed (%s)",
3517 iface->name, ovs_strerror(error));
3523 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3525 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3527 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3528 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3529 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3533 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3534 * sets the "ofport" field to -1.
3536 * This is appropriate when 'if_cfg''s interface cannot be created or is
3537 * otherwise invalid. */
3539 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3541 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3542 ovsrec_interface_set_status(if_cfg, NULL);
3543 ovsrec_interface_set_admin_state(if_cfg, NULL);
3544 ovsrec_interface_set_duplex(if_cfg, NULL);
3545 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3546 ovsrec_interface_set_link_state(if_cfg, NULL);
3547 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3548 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3549 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3550 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3551 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3552 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3553 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3554 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3559 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3561 union ovsdb_atom atom;
3563 atom.integer = target;
3564 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3568 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3570 struct ofpbuf queues_buf;
3572 ofpbuf_init(&queues_buf, 0);
3574 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3575 netdev_set_qos(iface->netdev, NULL, NULL);
3577 const struct ovsdb_datum *queues;
3578 struct netdev_queue_dump dump;
3579 unsigned int queue_id;
3580 struct smap details;
3584 /* Configure top-level Qos for 'iface'. */
3585 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3587 /* Deconfigure queues that were deleted. */
3588 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3590 smap_init(&details);
3591 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3592 if (!queue_ids_include(queues, queue_id)) {
3593 netdev_delete_queue(iface->netdev, queue_id);
3596 smap_destroy(&details);
3598 /* Configure queues for 'iface'. */
3600 for (i = 0; i < qos->n_queues; i++) {
3601 const struct ovsrec_queue *queue = qos->value_queues[i];
3602 unsigned int queue_id = qos->key_queues[i];
3604 if (queue_id == 0) {
3608 if (queue->n_dscp == 1) {
3609 struct ofproto_port_queue *port_queue;
3611 port_queue = ofpbuf_put_uninit(&queues_buf,
3612 sizeof *port_queue);
3613 port_queue->queue = queue_id;
3614 port_queue->dscp = queue->dscp[0];
3617 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3620 struct smap details;
3622 smap_init(&details);
3623 netdev_set_queue(iface->netdev, 0, &details);
3624 smap_destroy(&details);
3628 if (iface->ofp_port != OFPP_NONE) {
3629 const struct ofproto_port_queue *port_queues = queues_buf.data;
3630 size_t n_queues = queues_buf.size / sizeof *port_queues;
3632 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3633 port_queues, n_queues);
3636 netdev_set_policing(iface->netdev,
3637 iface->cfg->ingress_policing_rate,
3638 iface->cfg->ingress_policing_burst);
3640 ofpbuf_uninit(&queues_buf);
3644 iface_configure_cfm(struct iface *iface)
3646 const struct ovsrec_interface *cfg = iface->cfg;
3647 const char *opstate_str;
3648 const char *cfm_ccm_vlan;
3649 struct cfm_settings s;
3650 struct smap netdev_args;
3652 if (!cfg->n_cfm_mpid) {
3653 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3657 s.check_tnl_key = false;
3658 smap_init(&netdev_args);
3659 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3660 const char *key = smap_get(&netdev_args, "key");
3661 const char *in_key = smap_get(&netdev_args, "in_key");
3663 s.check_tnl_key = (key && !strcmp(key, "flow"))
3664 || (in_key && !strcmp(in_key, "flow"));
3666 smap_destroy(&netdev_args);
3668 s.mpid = *cfg->cfm_mpid;
3669 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3670 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3671 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3673 if (s.interval <= 0) {
3677 if (!cfm_ccm_vlan) {
3679 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3680 s.ccm_vlan = CFM_RANDOM_VLAN;
3682 s.ccm_vlan = atoi(cfm_ccm_vlan);
3683 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3688 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3690 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3692 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3693 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3695 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3698 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3699 * instead of obtaining it from the database. */
3701 iface_is_synthetic(const struct iface *iface)
3703 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3707 iface_validate_ofport__(size_t n, int64_t *ofport)
3709 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3710 ? u16_to_ofp(*ofport)
3715 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3717 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3721 iface_pick_ofport(const struct ovsrec_interface *cfg)
3723 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3724 return (requested_ofport != OFPP_NONE
3726 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3729 /* Port mirroring. */
3731 static struct mirror *
3732 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3736 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3737 if (uuid_equals(uuid, &m->uuid)) {
3745 bridge_configure_mirrors(struct bridge *br)
3747 const struct ovsdb_datum *mc;
3748 unsigned long *flood_vlans;
3749 struct mirror *m, *next;
3752 /* Get rid of deleted mirrors. */
3753 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3754 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3755 union ovsdb_atom atom;
3757 atom.uuid = m->uuid;
3758 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3763 /* Add new mirrors and reconfigure existing ones. */
3764 for (i = 0; i < br->cfg->n_mirrors; i++) {
3765 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3766 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3768 m = mirror_create(br, cfg);
3771 if (!mirror_configure(m)) {
3776 /* Update flooded vlans (for RSPAN). */
3777 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3778 br->cfg->n_flood_vlans);
3779 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3780 bitmap_free(flood_vlans);
3783 static struct mirror *
3784 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3788 m = xzalloc(sizeof *m);
3789 m->uuid = cfg->header_.uuid;
3790 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3792 m->name = xstrdup(cfg->name);
3798 mirror_destroy(struct mirror *m)
3801 struct bridge *br = m->bridge;
3804 ofproto_mirror_unregister(br->ofproto, m);
3807 hmap_remove(&br->mirrors, &m->hmap_node);
3814 mirror_collect_ports(struct mirror *m,
3815 struct ovsrec_port **in_ports, int n_in_ports,
3816 void ***out_portsp, size_t *n_out_portsp)
3818 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3819 size_t n_out_ports = 0;
3822 for (i = 0; i < n_in_ports; i++) {
3823 const char *name = in_ports[i]->name;
3824 struct port *port = port_lookup(m->bridge, name);
3826 out_ports[n_out_ports++] = port;
3828 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3829 "port %s", m->bridge->name, m->name, name);
3832 *out_portsp = out_ports;
3833 *n_out_portsp = n_out_ports;
3837 mirror_configure(struct mirror *m)
3839 const struct ovsrec_mirror *cfg = m->cfg;
3840 struct ofproto_mirror_settings s;
3843 if (strcmp(cfg->name, m->name)) {
3845 m->name = xstrdup(cfg->name);
3849 /* Get output port or VLAN. */
3850 if (cfg->output_port) {
3851 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3852 if (!s.out_bundle) {
3853 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3854 m->bridge->name, m->name);
3857 s.out_vlan = UINT16_MAX;
3859 if (cfg->output_vlan) {
3860 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3861 "output vlan; ignoring output vlan",
3862 m->bridge->name, m->name);
3864 } else if (cfg->output_vlan) {
3865 /* The database should prevent invalid VLAN values. */
3866 s.out_bundle = NULL;
3867 s.out_vlan = *cfg->output_vlan;
3869 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3870 m->bridge->name, m->name);
3874 /* Get port selection. */
3875 if (cfg->select_all) {
3876 size_t n_ports = hmap_count(&m->bridge->ports);
3877 void **ports = xmalloc(n_ports * sizeof *ports);
3882 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3892 /* Get ports, dropping ports that don't exist.
3893 * The IDL ensures that there are no duplicates. */
3894 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3895 &s.srcs, &s.n_srcs);
3896 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3897 &s.dsts, &s.n_dsts);
3900 /* Get VLAN selection. */
3901 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3904 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3907 if (s.srcs != s.dsts) {
3916 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3918 * This is deprecated. It is only for compatibility with broken device drivers
3919 * in old versions of Linux that do not properly support VLANs when VLAN
3920 * devices are not used. When broken device drivers are no longer in
3921 * widespread use, we will delete these interfaces. */
3923 static struct ovsrec_port **recs;
3924 static size_t n_recs, allocated_recs;
3926 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3927 * splinters are reconfigured. */
3929 register_rec(struct ovsrec_port *rec)
3931 if (n_recs >= allocated_recs) {
3932 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3934 recs[n_recs++] = rec;
3937 /* Frees all of the ports registered with register_reg(). */
3939 free_registered_recs(void)
3943 for (i = 0; i < n_recs; i++) {
3944 struct ovsrec_port *port = recs[i];
3947 for (j = 0; j < port->n_interfaces; j++) {
3948 struct ovsrec_interface *iface = port->interfaces[j];
3953 smap_destroy(&port->other_config);
3954 free(port->interfaces);
3962 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3965 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3967 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3971 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3974 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3975 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3976 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3979 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3980 * use, returns NULL.
3982 * Updates 'vlan_splinters_enabled_anywhere'. */
3983 static unsigned long int *
3984 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3986 unsigned long int *splinter_vlans;
3987 struct sset splinter_ifaces;
3988 const char *real_dev_name;
3989 struct shash *real_devs;
3990 struct shash_node *node;
3994 /* Free space allocated for synthesized ports and interfaces, since we're
3995 * in the process of reconstructing all of them. */
3996 free_registered_recs();
3998 splinter_vlans = bitmap_allocate(4096);
3999 sset_init(&splinter_ifaces);
4000 vlan_splinters_enabled_anywhere = false;
4001 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4002 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4005 for (j = 0; j < br_cfg->n_ports; j++) {
4006 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4009 for (k = 0; k < port_cfg->n_interfaces; k++) {
4010 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4012 if (vlan_splinters_is_enabled(iface_cfg)) {
4013 vlan_splinters_enabled_anywhere = true;
4014 sset_add(&splinter_ifaces, iface_cfg->name);
4015 vlan_bitmap_from_array__(port_cfg->trunks,
4021 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4022 bitmap_set1(splinter_vlans, *port_cfg->tag);
4027 if (!vlan_splinters_enabled_anywhere) {
4028 free(splinter_vlans);
4029 sset_destroy(&splinter_ifaces);
4033 HMAP_FOR_EACH (br, node, &all_bridges) {
4035 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4039 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4040 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4041 * device to be created for it. */
4042 bitmap_set0(splinter_vlans, 0);
4043 bitmap_set0(splinter_vlans, 4095);
4045 /* Delete all VLAN devices that we don't need. */
4047 real_devs = vlandev_get_real_devs();
4048 SHASH_FOR_EACH (node, real_devs) {
4049 const struct vlan_real_dev *real_dev = node->data;
4050 const struct vlan_dev *vlan_dev;
4051 bool real_dev_has_splinters;
4053 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4055 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4056 if (!real_dev_has_splinters
4057 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4058 struct netdev *netdev;
4060 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4061 if (!netdev_get_in4(netdev, NULL, NULL) ||
4062 !netdev_get_in6(netdev, NULL)) {
4063 /* It has an IP address configured, so we don't own
4064 * it. Don't delete it. */
4066 vlandev_del(vlan_dev->name);
4068 netdev_close(netdev);
4075 /* Add all VLAN devices that we need. */
4076 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4079 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4080 if (!vlandev_get_name(real_dev_name, vid)) {
4081 vlandev_add(real_dev_name, vid);
4088 sset_destroy(&splinter_ifaces);
4090 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4091 free(splinter_vlans);
4094 return splinter_vlans;
4097 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4100 configure_splinter_port(struct port *port)
4102 struct ofproto *ofproto = port->bridge->ofproto;
4103 ofp_port_t realdev_ofp_port;
4104 const char *realdev_name;
4105 struct iface *vlandev, *realdev;
4107 ofproto_bundle_unregister(port->bridge->ofproto, port);
4109 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4112 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4113 realdev = iface_lookup(port->bridge, realdev_name);
4114 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4116 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4120 static struct ovsrec_port *
4121 synthesize_splinter_port(const char *real_dev_name,
4122 const char *vlan_dev_name, int vid)
4124 struct ovsrec_interface *iface;
4125 struct ovsrec_port *port;
4127 iface = xmalloc(sizeof *iface);
4128 ovsrec_interface_init(iface);
4129 iface->name = xstrdup(vlan_dev_name);
4130 iface->type = "system";
4132 port = xmalloc(sizeof *port);
4133 ovsrec_port_init(port);
4134 port->interfaces = xmemdup(&iface, sizeof iface);
4135 port->n_interfaces = 1;
4136 port->name = xstrdup(vlan_dev_name);
4137 port->vlan_mode = "splinter";
4138 port->tag = xmalloc(sizeof *port->tag);
4141 smap_add(&port->other_config, "realdev", real_dev_name);
4147 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4148 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4149 * 1-bit in the 'splinter_vlans' bitmap. */
4151 add_vlan_splinter_ports(struct bridge *br,
4152 const unsigned long int *splinter_vlans,
4153 struct shash *ports)
4157 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4158 * we're modifying 'ports'. */
4159 for (i = 0; i < br->cfg->n_ports; i++) {
4160 const char *name = br->cfg->ports[i]->name;
4161 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4164 for (j = 0; j < port_cfg->n_interfaces; j++) {
4165 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4167 if (vlan_splinters_is_enabled(iface_cfg)) {
4168 const char *real_dev_name;
4171 real_dev_name = iface_cfg->name;
4172 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4173 const char *vlan_dev_name;
4175 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4177 && !shash_find(ports, vlan_dev_name)) {
4178 shash_add(ports, vlan_dev_name,
4179 synthesize_splinter_port(
4180 real_dev_name, vlan_dev_name, vid));
4189 mirror_refresh_stats(struct mirror *m)
4191 struct ofproto *ofproto = m->bridge->ofproto;
4192 uint64_t tx_packets, tx_bytes;
4195 size_t stat_cnt = 0;
4197 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4198 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4202 if (tx_packets != UINT64_MAX) {
4203 keys[stat_cnt] = "tx_packets";
4204 values[stat_cnt] = tx_packets;
4207 if (tx_bytes != UINT64_MAX) {
4208 keys[stat_cnt] = "tx_bytes";
4209 values[stat_cnt] = tx_bytes;
4213 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);