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.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
41 #include "ofproto/ofproto.h"
42 #include "poll-loop.h"
46 #include "socket-util.h"
48 #include "stream-ssl.h"
50 #include "system-stats.h"
55 #include "lib/vswitch-idl.h"
56 #include "xenserver.h"
58 #include "sflow_api.h"
59 #include "vlan-bitmap.h"
61 VLOG_DEFINE_THIS_MODULE(bridge);
63 COVERAGE_DEFINE(bridge_reconfigure);
65 /* Configuration of an uninstantiated iface. */
67 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
68 const struct ovsrec_interface *cfg; /* Interface record. */
69 const struct ovsrec_port *parent; /* Parent port record. */
70 ofp_port_t ofport; /* Requested OpenFlow port number. */
73 /* OpenFlow port slated for removal from ofproto. */
75 struct list list_node; /* Node in bridge's ofpp_garbage. */
76 ofp_port_t ofp_port; /* Port to be deleted. */
80 /* These members are always valid. */
81 struct list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
86 /* These members are valid only after bridge_reconfigure() causes them to
88 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
89 ofp_port_t ofp_port; /* OpenFlow port number, */
90 /* OFPP_NONE if unknown. */
91 struct netdev *netdev; /* Network device. */
92 const char *type; /* Usually same as cfg->type. */
93 const struct ovsrec_interface *cfg;
97 struct uuid uuid; /* UUID of this "mirror" record in database. */
98 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
99 struct bridge *bridge;
101 const struct ovsrec_mirror *cfg;
105 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
106 struct bridge *bridge;
109 const struct ovsrec_port *cfg;
111 /* An ordinary bridge port has 1 interface.
112 * A bridge port for bonding has at least 2 interfaces. */
113 struct list ifaces; /* List of "struct iface"s. */
117 struct hmap_node node; /* In 'all_bridges'. */
118 char *name; /* User-specified arbitrary name. */
119 char *type; /* Datapath type. */
120 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
121 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
122 const struct ovsrec_bridge *cfg;
124 /* OpenFlow switch processing. */
125 struct ofproto *ofproto; /* OpenFlow switch. */
128 struct hmap ports; /* "struct port"s indexed by name. */
129 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
130 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
132 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
133 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
134 Indexed on 'cfg->name'. */
136 /* Port mirroring. */
137 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
145 /* All bridges, indexed by name. */
146 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
148 /* OVSDB IDL used to obtain configuration. */
149 static struct ovsdb_idl *idl;
151 /* We want to complete daemonization, fully detaching from our parent process,
152 * only after we have completed our initial configuration, committed our state
153 * to the database, and received confirmation back from the database server
154 * that it applied the commit. This allows our parent process to know that,
155 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
156 * to have already been filled in. (It is only "very likely" rather than
157 * certain because there is always a slim possibility that the transaction will
158 * fail or that some other client has added new bridges, ports, etc. while
159 * ovs-vswitchd was configuring using an old configuration.)
161 * We only need to do this once for our initial configuration at startup, so
162 * 'initial_config_done' tracks whether we've already done it. While we are
163 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
164 * itself and is otherwise NULL. */
165 static bool initial_config_done;
166 static struct ovsdb_idl_txn *daemonize_txn;
168 /* Most recently processed IDL sequence number. */
169 static unsigned int idl_seqno;
171 /* Each time this timer expires, the bridge fetches interface and mirror
172 * statistics and pushes them into the database. */
173 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
174 static long long int iface_stats_timer = LLONG_MIN;
176 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
177 * expensive. This can cause bridge_reconfigure() to take a long time during
178 * which no other work can be done. To deal with this problem, we limit port
179 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
180 * call to bridge_reconfigure(). If there is more work to do after the limit
181 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
182 * This allows the rest of the code to catch up on important things like
183 * forwarding packets. */
184 #define OFP_PORT_ACTION_WINDOW 10
185 static bool reconfiguring = false;
187 static void add_del_bridges(const struct ovsrec_open_vswitch *);
188 static void bridge_update_ofprotos(void);
189 static void bridge_create(const struct ovsrec_bridge *);
190 static void bridge_destroy(struct bridge *);
191 static struct bridge *bridge_lookup(const char *name);
192 static unixctl_cb_func bridge_unixctl_dump_flows;
193 static unixctl_cb_func bridge_unixctl_reconnect;
194 static size_t bridge_get_controllers(const struct bridge *br,
195 struct ovsrec_controller ***controllersp);
196 static void bridge_add_del_ports(struct bridge *,
197 const unsigned long int *splinter_vlans);
198 static void bridge_refresh_ofp_port(struct bridge *);
199 static void bridge_configure_datapath_id(struct bridge *);
200 static void bridge_configure_netflow(struct bridge *);
201 static void bridge_configure_forward_bpdu(struct bridge *);
202 static void bridge_configure_mac_table(struct bridge *);
203 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
204 static void bridge_configure_ipfix(struct bridge *);
205 static void bridge_configure_stp(struct bridge *);
206 static void bridge_configure_tables(struct bridge *);
207 static void bridge_configure_dp_desc(struct bridge *);
208 static void bridge_configure_remotes(struct bridge *,
209 const struct sockaddr_in *managers,
211 static void bridge_pick_local_hw_addr(struct bridge *,
212 uint8_t ea[ETH_ADDR_LEN],
213 struct iface **hw_addr_iface);
214 static uint64_t bridge_pick_datapath_id(struct bridge *,
215 const uint8_t bridge_ea[ETH_ADDR_LEN],
216 struct iface *hw_addr_iface);
217 static void bridge_queue_if_cfg(struct bridge *,
218 const struct ovsrec_interface *,
219 const struct ovsrec_port *);
220 static uint64_t dpid_from_hash(const void *, size_t nbytes);
221 static bool bridge_has_bond_fake_iface(const struct bridge *,
223 static bool port_is_bond_fake_iface(const struct port *);
225 static unixctl_cb_func qos_unixctl_show;
227 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
228 static void port_del_ifaces(struct port *);
229 static void port_destroy(struct port *);
230 static struct port *port_lookup(const struct bridge *, const char *name);
231 static void port_configure(struct port *);
232 static struct lacp_settings *port_configure_lacp(struct port *,
233 struct lacp_settings *);
234 static void port_configure_bond(struct port *, struct bond_settings *);
235 static bool port_is_synthetic(const struct port *);
237 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
238 static void run_system_stats(void);
240 static void bridge_configure_mirrors(struct bridge *);
241 static struct mirror *mirror_create(struct bridge *,
242 const struct ovsrec_mirror *);
243 static void mirror_destroy(struct mirror *);
244 static bool mirror_configure(struct mirror *);
245 static void mirror_refresh_stats(struct mirror *);
247 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
248 static bool iface_create(struct bridge *, struct if_cfg *,
249 ofp_port_t ofp_port);
250 static bool iface_is_internal(const struct ovsrec_interface *iface,
251 const struct ovsrec_bridge *br);
252 static const char *iface_get_type(const struct ovsrec_interface *,
253 const struct ovsrec_bridge *);
254 static void iface_destroy(struct iface *);
255 static struct iface *iface_lookup(const struct bridge *, const char *name);
256 static struct iface *iface_find(const char *name);
257 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_from_ofp_port(const struct bridge *,
259 ofp_port_t ofp_port);
260 static void iface_set_mac(struct iface *);
261 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
262 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
263 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
264 static void iface_configure_cfm(struct iface *);
265 static void iface_refresh_cfm_stats(struct iface *);
266 static void iface_refresh_stats(struct iface *);
267 static void iface_refresh_status(struct iface *);
268 static bool iface_is_synthetic(const struct iface *);
269 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
271 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
273 * This is deprecated. It is only for compatibility with broken device drivers
274 * in old versions of Linux that do not properly support VLANs when VLAN
275 * devices are not used. When broken device drivers are no longer in
276 * widespread use, we will delete these interfaces. */
278 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
279 static bool vlan_splinters_enabled_anywhere;
281 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
282 static unsigned long int *collect_splinter_vlans(
283 const struct ovsrec_open_vswitch *);
284 static void configure_splinter_port(struct port *);
285 static void add_vlan_splinter_ports(struct bridge *,
286 const unsigned long int *splinter_vlans,
287 struct shash *ports);
290 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
292 struct shash iface_hints;
293 static bool initialized = false;
300 shash_init(&iface_hints);
303 for (i = 0; i < cfg->n_bridges; i++) {
304 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
307 for (j = 0; j < br_cfg->n_ports; j++) {
308 struct ovsrec_port *port_cfg = br_cfg->ports[j];
311 for (k = 0; k < port_cfg->n_interfaces; k++) {
312 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
313 struct iface_hint *iface_hint;
315 iface_hint = xmalloc(sizeof *iface_hint);
316 iface_hint->br_name = br_cfg->name;
317 iface_hint->br_type = br_cfg->datapath_type;
318 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
320 shash_add(&iface_hints, if_cfg->name, iface_hint);
326 ofproto_init(&iface_hints);
328 shash_destroy_free_data(&iface_hints);
332 /* Public functions. */
334 /* Initializes the bridge module, configuring it to obtain its configuration
335 * from an OVSDB server accessed over 'remote', which should be a string in a
336 * form acceptable to ovsdb_idl_create(). */
338 bridge_init(const char *remote)
340 /* Create connection to database. */
341 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
342 idl_seqno = ovsdb_idl_get_seqno(idl);
343 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
344 ovsdb_idl_verify_write_only(idl);
346 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
347 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
354 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
355 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
356 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
358 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
359 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
360 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
361 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
380 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
385 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
392 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
394 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
405 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
407 /* Register unixctl commands. */
408 unixctl_command_register("qos/show", "interface", 1, 1,
409 qos_unixctl_show, NULL);
410 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
411 bridge_unixctl_dump_flows, NULL);
412 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
413 bridge_unixctl_reconnect, NULL);
423 struct bridge *br, *next_br;
425 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
428 ovsdb_idl_destroy(idl);
431 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
432 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
433 * responsible for freeing '*managersp' (with free()).
435 * You may be asking yourself "why does ovs-vswitchd care?", because
436 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
437 * should not be and in fact is not directly involved in that. But
438 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
439 * it has to tell in-band control where the managers are to enable that.
440 * (Thus, only managers connected in-band are collected.)
443 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
444 struct sockaddr_in **managersp, size_t *n_managersp)
446 struct sockaddr_in *managers = NULL;
447 size_t n_managers = 0;
451 /* Collect all of the potential targets from the "targets" columns of the
452 * rows pointed to by "manager_options", excluding any that are
455 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
456 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
458 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
459 sset_find_and_delete(&targets, m->target);
461 sset_add(&targets, m->target);
465 /* Now extract the targets' IP addresses. */
466 if (!sset_is_empty(&targets)) {
469 managers = xmalloc(sset_count(&targets) * sizeof *managers);
470 SSET_FOR_EACH (target, &targets) {
471 struct sockaddr_in *sin = &managers[n_managers];
473 if (stream_parse_target_with_default_ports(target,
481 sset_destroy(&targets);
483 *managersp = managers;
484 *n_managersp = n_managers;
488 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
490 unsigned long int *splinter_vlans;
493 COVERAGE_INC(bridge_reconfigure);
495 ovs_assert(!reconfiguring);
496 reconfiguring = true;
498 ofproto_set_flow_eviction_threshold(
499 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
500 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
502 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
503 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
504 * configuration otherwise.
506 * This is mostly an update to bridge data structures. Nothing is pushed
507 * down to ofproto or lower layers. */
508 add_del_bridges(ovs_cfg);
509 splinter_vlans = collect_splinter_vlans(ovs_cfg);
510 HMAP_FOR_EACH (br, node, &all_bridges) {
511 bridge_add_del_ports(br, splinter_vlans);
513 free(splinter_vlans);
515 /* Delete datapaths that are no longer configured, and create ones which
516 * don't exist but should. */
517 bridge_update_ofprotos();
519 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
520 HMAP_FOR_EACH (br, node, &all_bridges) {
521 bridge_refresh_ofp_port(br);
524 /* Clear database records for "if_cfg"s which haven't been instantiated. */
525 HMAP_FOR_EACH (br, node, &all_bridges) {
526 struct if_cfg *if_cfg;
528 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
529 iface_clear_db_record(if_cfg->cfg);
533 reconfigure_system_stats(ovs_cfg);
537 bridge_reconfigure_ofp(void)
539 long long int deadline;
543 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
545 /* The kernel will reject any attempt to add a given port to a datapath if
546 * that port already belongs to a different datapath, so we must do all
547 * port deletions before any port additions. */
548 HMAP_FOR_EACH (br, node, &all_bridges) {
549 struct ofpp_garbage *garbage, *next;
551 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
552 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
553 * internal datastructures may not be consistent. Eventually, when
554 * port additions and deletions are cheaper, these calls should be
557 ofproto_port_del(br->ofproto, garbage->ofp_port);
558 list_remove(&garbage->list_node);
562 if (time_msec() >= deadline) {
569 HMAP_FOR_EACH (br, node, &all_bridges) {
570 struct if_cfg *if_cfg, *next;
572 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
573 iface_create(br, if_cfg, OFPP_NONE);
575 if (time_msec() >= deadline) {
585 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
587 struct sockaddr_in *managers;
588 int sflow_bridge_number;
593 ovs_assert(reconfiguring);
594 done = bridge_reconfigure_ofp();
596 /* Complete the configuration. */
597 sflow_bridge_number = 0;
598 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
599 HMAP_FOR_EACH (br, node, &all_bridges) {
602 /* We need the datapath ID early to allow LACP ports to use it as the
603 * default system ID. */
604 bridge_configure_datapath_id(br);
606 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
609 port_configure(port);
611 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
612 iface_configure_cfm(iface);
613 iface_configure_qos(iface, port->cfg->qos);
614 iface_set_mac(iface);
615 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
619 bridge_configure_mirrors(br);
620 bridge_configure_forward_bpdu(br);
621 bridge_configure_mac_table(br);
622 bridge_configure_remotes(br, managers, n_managers);
623 bridge_configure_netflow(br);
624 bridge_configure_sflow(br, &sflow_bridge_number);
625 bridge_configure_ipfix(br);
626 bridge_configure_stp(br);
627 bridge_configure_tables(br);
628 bridge_configure_dp_desc(br);
630 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
631 /* XXX: Remove this warning message eventually. */
632 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
633 " moved to the Open_vSwitch table. Ignoring its"
634 " setting in the bridge table.");
642 /* Delete ofprotos which aren't configured or have the wrong type. Create
643 * ofprotos which don't exist but need to. */
645 bridge_update_ofprotos(void)
647 struct bridge *br, *next;
652 /* Delete ofprotos with no bridge or with the wrong type. */
655 ofproto_enumerate_types(&types);
656 SSET_FOR_EACH (type, &types) {
659 ofproto_enumerate_names(type, &names);
660 SSET_FOR_EACH (name, &names) {
661 br = bridge_lookup(name);
662 if (!br || strcmp(type, br->type)) {
663 ofproto_delete(name, type);
667 sset_destroy(&names);
668 sset_destroy(&types);
670 /* Add ofprotos for bridges which don't have one yet. */
671 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
679 /* Remove ports from any datapath with the same name as 'br'. If we
680 * don't do this, creating 'br''s ofproto will fail because a port with
681 * the same name as its local port already exists. */
682 HMAP_FOR_EACH (br2, node, &all_bridges) {
683 struct ofproto_port ofproto_port;
689 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
691 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
693 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
696 ofproto_port_destroy(&ofproto_port);
700 error = ofproto_create(br->name, br->type, &br->ofproto);
702 VLOG_ERR("failed to create bridge %s: %s", br->name,
710 port_configure(struct port *port)
712 const struct ovsrec_port *cfg = port->cfg;
713 struct bond_settings bond_settings;
714 struct lacp_settings lacp_settings;
715 struct ofproto_bundle_settings s;
718 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
719 configure_splinter_port(port);
728 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
729 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
730 s.slaves[s.n_slaves++] = iface->ofp_port;
735 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
739 /* Get VLAN trunks. */
742 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
746 if (cfg->vlan_mode) {
747 if (!strcmp(cfg->vlan_mode, "access")) {
748 s.vlan_mode = PORT_VLAN_ACCESS;
749 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
750 s.vlan_mode = PORT_VLAN_TRUNK;
751 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
752 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
753 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
754 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
756 /* This "can't happen" because ovsdb-server should prevent it. */
757 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
758 s.vlan_mode = PORT_VLAN_TRUNK;
762 s.vlan_mode = PORT_VLAN_ACCESS;
764 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
768 s.vlan_mode = PORT_VLAN_TRUNK;
771 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
774 /* Get LACP settings. */
775 s.lacp = port_configure_lacp(port, &lacp_settings);
779 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
780 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
781 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
784 s.lacp_slaves = NULL;
787 /* Get bond settings. */
788 if (s.n_slaves > 1) {
789 s.bond = &bond_settings;
790 port_configure_bond(port, &bond_settings);
793 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
794 netdev_set_miimon_interval(iface->netdev, 0);
799 ofproto_bundle_register(port->bridge->ofproto, port, &s);
807 /* Pick local port hardware address and datapath ID for 'br'. */
809 bridge_configure_datapath_id(struct bridge *br)
811 uint8_t ea[ETH_ADDR_LEN];
813 struct iface *local_iface;
814 struct iface *hw_addr_iface;
817 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
818 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
820 int error = netdev_set_etheraddr(local_iface->netdev, ea);
822 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
823 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
824 "Ethernet address: %s",
825 br->name, strerror(error));
828 memcpy(br->ea, ea, ETH_ADDR_LEN);
830 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
831 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
832 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
833 ofproto_set_datapath_id(br->ofproto, dpid);
836 dpid_string = xasprintf("%016"PRIx64, dpid);
837 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
841 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
844 bridge_get_allowed_versions(struct bridge *br)
846 if (!br->cfg->n_protocols)
849 return ofputil_versions_from_strings(br->cfg->protocols,
850 br->cfg->n_protocols);
853 /* Set NetFlow configuration on 'br'. */
855 bridge_configure_netflow(struct bridge *br)
857 struct ovsrec_netflow *cfg = br->cfg->netflow;
858 struct netflow_options opts;
861 ofproto_set_netflow(br->ofproto, NULL);
865 memset(&opts, 0, sizeof opts);
867 /* Get default NetFlow configuration from datapath.
868 * Apply overrides from 'cfg'. */
869 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
870 if (cfg->engine_type) {
871 opts.engine_type = *cfg->engine_type;
873 if (cfg->engine_id) {
874 opts.engine_id = *cfg->engine_id;
877 /* Configure active timeout interval. */
878 opts.active_timeout = cfg->active_timeout;
879 if (!opts.active_timeout) {
880 opts.active_timeout = -1;
881 } else if (opts.active_timeout < 0) {
882 VLOG_WARN("bridge %s: active timeout interval set to negative "
883 "value, using default instead (%d seconds)", br->name,
884 NF_ACTIVE_TIMEOUT_DEFAULT);
885 opts.active_timeout = -1;
888 /* Add engine ID to interface number to disambiguate bridgs? */
889 opts.add_id_to_iface = cfg->add_id_to_interface;
890 if (opts.add_id_to_iface) {
891 if (opts.engine_id > 0x7f) {
892 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
893 "another vswitch, choose an engine id less than 128",
896 if (hmap_count(&br->ports) > 508) {
897 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
898 "another port when more than 508 ports are used",
904 sset_init(&opts.collectors);
905 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
908 if (ofproto_set_netflow(br->ofproto, &opts)) {
909 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
911 sset_destroy(&opts.collectors);
914 /* Set sFlow configuration on 'br'. */
916 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
918 const struct ovsrec_sflow *cfg = br->cfg->sflow;
919 struct ovsrec_controller **controllers;
920 struct ofproto_sflow_options oso;
921 size_t n_controllers;
925 ofproto_set_sflow(br->ofproto, NULL);
929 memset(&oso, 0, sizeof oso);
931 sset_init(&oso.targets);
932 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
934 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
936 oso.sampling_rate = *cfg->sampling;
939 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
941 oso.polling_interval = *cfg->polling;
944 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
946 oso.header_len = *cfg->header;
949 oso.sub_id = (*sflow_bridge_number)++;
950 oso.agent_device = cfg->agent;
952 oso.control_ip = NULL;
953 n_controllers = bridge_get_controllers(br, &controllers);
954 for (i = 0; i < n_controllers; i++) {
955 if (controllers[i]->local_ip) {
956 oso.control_ip = controllers[i]->local_ip;
960 ofproto_set_sflow(br->ofproto, &oso);
962 sset_destroy(&oso.targets);
965 /* Set IPFIX configuration on 'br'. */
967 bridge_configure_ipfix(struct bridge *br)
969 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
970 const struct ovsrec_flow_sample_collector_set *fe_cfg;
971 struct ofproto_ipfix_bridge_exporter_options be_opts;
972 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
973 size_t n_fe_opts = 0;
975 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
976 if (fe_cfg->bridge == br->cfg) {
981 if (!be_cfg && n_fe_opts == 0) {
982 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
987 memset(&be_opts, 0, sizeof be_opts);
989 sset_init(&be_opts.targets);
990 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
992 if (be_cfg->sampling) {
993 be_opts.sampling_rate = *be_cfg->sampling;
995 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
997 if (be_cfg->obs_domain_id) {
998 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1000 if (be_cfg->obs_point_id) {
1001 be_opts.obs_point_id = *be_cfg->obs_point_id;
1005 if (n_fe_opts > 0) {
1006 struct ofproto_ipfix_flow_exporter_options *opts;
1007 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1009 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1010 if (fe_cfg->bridge == br->cfg) {
1011 opts->collector_set_id = fe_cfg->id;
1012 sset_init(&opts->targets);
1013 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1014 fe_cfg->ipfix->n_targets);
1020 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1024 sset_destroy(&be_opts.targets);
1027 if (n_fe_opts > 0) {
1028 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1030 for (i = 0; i < n_fe_opts; i++) {
1031 sset_destroy(&opts->targets);
1039 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1040 struct ofproto_port_stp_settings *port_s,
1041 int *port_num_counter, unsigned long *port_num_bitmap)
1043 const char *config_str;
1044 struct iface *iface;
1046 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1047 port_s->enable = false;
1050 port_s->enable = true;
1053 /* STP over bonds is not supported. */
1054 if (!list_is_singleton(&port->ifaces)) {
1055 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1057 port_s->enable = false;
1061 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1063 /* Internal ports shouldn't participate in spanning tree, so
1065 if (!strcmp(iface->type, "internal")) {
1066 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1067 port_s->enable = false;
1071 /* STP on mirror output ports is not supported. */
1072 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1073 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1074 port_s->enable = false;
1078 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1080 unsigned long int port_num = strtoul(config_str, NULL, 0);
1081 int port_idx = port_num - 1;
1083 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1084 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1085 port_s->enable = false;
1089 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1090 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1091 port->name, port_num);
1092 port_s->enable = false;
1095 bitmap_set1(port_num_bitmap, port_idx);
1096 port_s->port_num = port_idx;
1098 if (*port_num_counter >= STP_MAX_PORTS) {
1099 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1100 port_s->enable = false;
1104 port_s->port_num = (*port_num_counter)++;
1107 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1109 port_s->path_cost = strtoul(config_str, NULL, 10);
1111 enum netdev_features current;
1114 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1115 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1116 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1119 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1121 port_s->priority = strtoul(config_str, NULL, 0);
1123 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1127 /* Set spanning tree configuration on 'br'. */
1129 bridge_configure_stp(struct bridge *br)
1131 if (!br->cfg->stp_enable) {
1132 ofproto_set_stp(br->ofproto, NULL);
1134 struct ofproto_stp_settings br_s;
1135 const char *config_str;
1137 int port_num_counter;
1138 unsigned long *port_num_bitmap;
1140 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1142 uint8_t ea[ETH_ADDR_LEN];
1144 if (eth_addr_from_string(config_str, ea)) {
1145 br_s.system_id = eth_addr_to_uint64(ea);
1147 br_s.system_id = eth_addr_to_uint64(br->ea);
1148 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1149 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1152 br_s.system_id = eth_addr_to_uint64(br->ea);
1155 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1157 br_s.priority = strtoul(config_str, NULL, 0);
1159 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1162 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1164 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1166 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1169 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1171 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1173 br_s.max_age = STP_DEFAULT_MAX_AGE;
1176 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1178 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1180 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1183 /* Configure STP on the bridge. */
1184 if (ofproto_set_stp(br->ofproto, &br_s)) {
1185 VLOG_ERR("bridge %s: could not enable STP", br->name);
1189 /* Users must either set the port number with the "stp-port-num"
1190 * configuration on all ports or none. If manual configuration
1191 * is not done, then we allocate them sequentially. */
1192 port_num_counter = 0;
1193 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1194 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1195 struct ofproto_port_stp_settings port_s;
1196 struct iface *iface;
1198 port_configure_stp(br->ofproto, port, &port_s,
1199 &port_num_counter, port_num_bitmap);
1201 /* As bonds are not supported, just apply configuration to
1202 * all interfaces. */
1203 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1204 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1206 VLOG_ERR("port %s: could not enable STP", port->name);
1212 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1213 && port_num_counter) {
1214 VLOG_ERR("bridge %s: must manually configure all STP port "
1215 "IDs or none, disabling", br->name);
1216 ofproto_set_stp(br->ofproto, NULL);
1218 bitmap_free(port_num_bitmap);
1223 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1225 const struct port *port = port_lookup(br, name);
1226 return port && port_is_bond_fake_iface(port);
1230 port_is_bond_fake_iface(const struct port *port)
1232 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1236 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1238 struct bridge *br, *next;
1239 struct shash new_br;
1242 /* Collect new bridges' names and types. */
1243 shash_init(&new_br);
1244 for (i = 0; i < cfg->n_bridges; i++) {
1245 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1246 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1248 if (strchr(br_cfg->name, '/')) {
1249 /* Prevent remote ovsdb-server users from accessing arbitrary
1250 * directories, e.g. consider a bridge named "../../../etc/". */
1251 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1253 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1254 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1258 /* Get rid of deleted bridges or those whose types have changed.
1259 * Update 'cfg' of bridges that still exist. */
1260 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1261 br->cfg = shash_find_data(&new_br, br->name);
1262 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1263 br->cfg->datapath_type))) {
1268 /* Add new bridges. */
1269 for (i = 0; i < cfg->n_bridges; i++) {
1270 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1271 struct bridge *br = bridge_lookup(br_cfg->name);
1273 bridge_create(br_cfg);
1277 shash_destroy(&new_br);
1281 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1283 struct bridge *br = iface->port->bridge;
1285 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1286 iface->ofp_port = ofp_port;
1287 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1288 hash_int(ofp_to_u16(ofp_port), 0));
1289 iface_set_ofport(iface->cfg, ofp_port);
1292 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1293 * Returns 0 if successful, otherwise a positive errno value. */
1295 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1296 struct netdev *netdev)
1300 error = netdev_set_config(netdev, &iface_cfg->options);
1302 VLOG_WARN("could not configure network device %s (%s)",
1303 iface_cfg->name, strerror(error));
1308 /* This function determines whether 'ofproto_port', which is attached to
1309 * br->ofproto's datapath, is one that we want in 'br'.
1311 * If it is, it returns true, after creating an iface (if necessary),
1312 * configuring the iface's netdev according to the iface's options, and setting
1313 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1315 * If, on the other hand, 'port' should be removed, it returns false. The
1316 * caller should later detach the port from br->ofproto. */
1318 bridge_refresh_one_ofp_port(struct bridge *br,
1319 const struct ofproto_port *ofproto_port)
1321 const char *name = ofproto_port->name;
1322 const char *type = ofproto_port->type;
1323 ofp_port_t ofp_port = ofproto_port->ofp_port;
1325 struct iface *iface = iface_lookup(br, name);
1327 /* Check that the name-to-number mapping is one-to-one. */
1328 if (iface->ofp_port != OFPP_NONE) {
1329 VLOG_WARN("bridge %s: interface %s reported twice",
1332 } else if (iface_from_ofp_port(br, ofp_port)) {
1333 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1334 br->name, ofp_port);
1338 /* There's a configured interface named 'name'. */
1339 if (strcmp(type, iface->type)
1340 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1341 /* It's the wrong type, or it's the right type but can't be
1342 * configured as the user requested, so we must destroy it. */
1345 /* It's the right type and configured correctly. Keep it. */
1346 iface_set_ofp_port(iface, ofp_port);
1349 } else if (bridge_has_bond_fake_iface(br, name)
1350 && !strcmp(type, "internal")) {
1351 /* It's a bond fake iface. Keep it. */
1354 /* There's no configured interface named 'name', but there might be an
1355 * interface of that name queued to be created.
1357 * If there is, and it has the correct type, then try to configure it
1358 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1359 * delete it and later try to re-add it. */
1360 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1362 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1363 && iface_create(br, if_cfg, ofp_port));
1367 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1368 * consistent with the ofp_ports in "br->ofproto". */
1370 bridge_refresh_ofp_port(struct bridge *br)
1372 struct ofproto_port_dump dump;
1373 struct ofproto_port ofproto_port;
1374 struct port *port, *port_next;
1376 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1377 hmap_clear(&br->ifaces);
1378 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1379 struct iface *iface;
1381 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1382 iface->ofp_port = OFPP_NONE;
1386 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1387 * already exist in the datapath and promote them to full fledged "struct
1388 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1389 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1390 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1391 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1392 garbage->ofp_port = ofproto_port.ofp_port;
1393 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1397 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1398 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1399 * later they can be added to ofproto. */
1400 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1401 struct iface *iface, *iface_next;
1403 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1404 if (iface->ofp_port == OFPP_NONE) {
1405 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1406 iface_destroy(iface);
1410 if (list_is_empty(&port->ifaces)) {
1416 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1417 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1418 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1421 * If successful, returns 0 and stores the network device in '*netdevp'. On
1422 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1424 iface_do_create(const struct bridge *br,
1425 const struct if_cfg *if_cfg,
1426 ofp_port_t *ofp_portp, struct netdev **netdevp)
1428 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1429 const struct ovsrec_port *port_cfg = if_cfg->parent;
1430 struct netdev *netdev = NULL;
1433 if (netdev_is_reserved_name(iface_cfg->name)) {
1434 VLOG_WARN("could not create interface %s, name is reserved",
1440 error = netdev_open(iface_cfg->name,
1441 iface_get_type(iface_cfg, br->cfg), &netdev);
1443 VLOG_WARN("could not open network device %s (%s)",
1444 iface_cfg->name, strerror(error));
1448 error = iface_set_netdev_config(iface_cfg, netdev);
1453 if (*ofp_portp == OFPP_NONE) {
1454 ofp_port_t ofp_port = if_cfg->ofport;
1456 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1460 *ofp_portp = ofp_port;
1462 VLOG_INFO("bridge %s: added interface %s on port %d",
1463 br->name, iface_cfg->name, *ofp_portp);
1465 VLOG_DBG("bridge %s: interface %s is on port %d",
1466 br->name, iface_cfg->name, *ofp_portp);
1469 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1470 || iface_is_internal(iface_cfg, br->cfg)) {
1471 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1479 netdev_close(netdev);
1483 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1484 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1485 * automatically allocated for the iface. Takes ownership of and
1486 * deallocates 'if_cfg'.
1488 * Return true if an iface is successfully created, false otherwise. */
1490 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1492 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1493 const struct ovsrec_port *port_cfg = if_cfg->parent;
1495 struct netdev *netdev;
1496 struct iface *iface;
1501 /* Do the bits that can fail up front.
1503 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1504 * internal datastructures may not be consistent. Eventually, when port
1505 * additions and deletions are cheaper, these calls should be removed. */
1507 ovs_assert(!iface_lookup(br, iface_cfg->name));
1508 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1511 iface_set_ofport(iface_cfg, OFPP_NONE);
1512 iface_clear_db_record(iface_cfg);
1517 /* Get or create the port structure. */
1518 port = port_lookup(br, port_cfg->name);
1520 port = port_create(br, port_cfg);
1523 /* Create the iface structure. */
1524 iface = xzalloc(sizeof *iface);
1525 list_push_back(&port->ifaces, &iface->port_elem);
1526 hmap_insert(&br->iface_by_name, &iface->name_node,
1527 hash_string(iface_cfg->name, 0));
1529 iface->name = xstrdup(iface_cfg->name);
1530 iface->ofp_port = OFPP_NONE;
1531 iface->netdev = netdev;
1532 iface->type = iface_get_type(iface_cfg, br->cfg);
1533 iface->cfg = iface_cfg;
1535 iface_set_ofp_port(iface, ofp_port);
1537 /* Populate initial status in database. */
1538 iface_refresh_stats(iface);
1539 iface_refresh_status(iface);
1541 /* Add bond fake iface if necessary. */
1542 if (port_is_bond_fake_iface(port)) {
1543 struct ofproto_port ofproto_port;
1545 if (ofproto_port_query_by_name(br->ofproto, port->name,
1547 struct netdev *netdev;
1550 error = netdev_open(port->name, "internal", &netdev);
1552 ofp_port_t fake_ofp_port = if_cfg->ofport;
1554 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1555 netdev_close(netdev);
1557 VLOG_WARN("could not open network device %s (%s)",
1558 port->name, strerror(error));
1561 /* Already exists, nothing to do. */
1562 ofproto_port_destroy(&ofproto_port);
1567 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1573 /* Set forward BPDU option. */
1575 bridge_configure_forward_bpdu(struct bridge *br)
1577 ofproto_set_forward_bpdu(br->ofproto,
1578 smap_get_bool(&br->cfg->other_config,
1583 /* Set MAC learning table configuration for 'br'. */
1585 bridge_configure_mac_table(struct bridge *br)
1587 const char *idle_time_str;
1590 const char *mac_table_size_str;
1593 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1594 idle_time = (idle_time_str && atoi(idle_time_str)
1595 ? atoi(idle_time_str)
1596 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1598 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1599 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1600 ? atoi(mac_table_size_str)
1603 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1607 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1608 struct iface **hw_addr_iface)
1610 struct hmapx mirror_output_ports;
1613 bool found_addr = false;
1617 *hw_addr_iface = NULL;
1619 /* Did the user request a particular MAC? */
1620 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1621 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1622 if (eth_addr_is_multicast(ea)) {
1623 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1624 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1625 } else if (eth_addr_is_zero(ea)) {
1626 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1632 /* Mirror output ports don't participate in picking the local hardware
1633 * address. ofproto can't help us find out whether a given port is a
1634 * mirror output because we haven't configured mirrors yet, so we need to
1635 * accumulate them ourselves. */
1636 hmapx_init(&mirror_output_ports);
1637 for (i = 0; i < br->cfg->n_mirrors; i++) {
1638 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1639 if (m->output_port) {
1640 hmapx_add(&mirror_output_ports, m->output_port);
1644 /* Otherwise choose the minimum non-local MAC address among all of the
1646 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1647 uint8_t iface_ea[ETH_ADDR_LEN];
1648 struct iface *candidate;
1649 struct iface *iface;
1651 /* Mirror output ports don't participate. */
1652 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1656 /* Choose the MAC address to represent the port. */
1658 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1659 /* Find the interface with this Ethernet address (if any) so that
1660 * we can provide the correct devname to the caller. */
1661 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1662 uint8_t candidate_ea[ETH_ADDR_LEN];
1663 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1664 && eth_addr_equals(iface_ea, candidate_ea)) {
1669 /* Choose the interface whose MAC address will represent the port.
1670 * The Linux kernel bonding code always chooses the MAC address of
1671 * the first slave added to a bond, and the Fedora networking
1672 * scripts always add slaves to a bond in alphabetical order, so
1673 * for compatibility we choose the interface with the name that is
1674 * first in alphabetical order. */
1675 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1676 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1681 /* The local port doesn't count (since we're trying to choose its
1682 * MAC address anyway). */
1683 if (iface->ofp_port == OFPP_LOCAL) {
1688 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1694 /* Compare against our current choice. */
1695 if (!eth_addr_is_multicast(iface_ea) &&
1696 !eth_addr_is_local(iface_ea) &&
1697 !eth_addr_is_reserved(iface_ea) &&
1698 !eth_addr_is_zero(iface_ea) &&
1699 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1701 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1702 *hw_addr_iface = iface;
1708 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1709 *hw_addr_iface = NULL;
1712 hmapx_destroy(&mirror_output_ports);
1715 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1716 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1717 * an interface on 'br', then that interface must be passed in as
1718 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1719 * 'hw_addr_iface' must be passed in as a null pointer. */
1721 bridge_pick_datapath_id(struct bridge *br,
1722 const uint8_t bridge_ea[ETH_ADDR_LEN],
1723 struct iface *hw_addr_iface)
1726 * The procedure for choosing a bridge MAC address will, in the most
1727 * ordinary case, also choose a unique MAC that we can use as a datapath
1728 * ID. In some special cases, though, multiple bridges will end up with
1729 * the same MAC address. This is OK for the bridges, but it will confuse
1730 * the OpenFlow controller, because each datapath needs a unique datapath
1733 * Datapath IDs must be unique. It is also very desirable that they be
1734 * stable from one run to the next, so that policy set on a datapath
1737 const char *datapath_id;
1740 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1741 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1745 if (!hw_addr_iface) {
1747 * A purely internal bridge, that is, one that has no non-virtual
1748 * network devices on it at all, is difficult because it has no
1749 * natural unique identifier at all.
1751 * When the host is a XenServer, we handle this case by hashing the
1752 * host's UUID with the name of the bridge. Names of bridges are
1753 * persistent across XenServer reboots, although they can be reused if
1754 * an internal network is destroyed and then a new one is later
1755 * created, so this is fairly effective.
1757 * When the host is not a XenServer, we punt by using a random MAC
1758 * address on each run.
1760 const char *host_uuid = xenserver_get_host_uuid();
1762 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1763 dpid = dpid_from_hash(combined, strlen(combined));
1769 return eth_addr_to_uint64(bridge_ea);
1773 dpid_from_hash(const void *data, size_t n)
1775 uint8_t hash[SHA1_DIGEST_SIZE];
1777 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1778 sha1_bytes(data, n, hash);
1779 eth_addr_mark_random(hash);
1780 return eth_addr_to_uint64(hash);
1784 iface_refresh_status(struct iface *iface)
1788 enum netdev_features current;
1792 uint8_t mac[ETH_ADDR_LEN];
1795 if (iface_is_synthetic(iface)) {
1801 if (!netdev_get_status(iface->netdev, &smap)) {
1802 ovsrec_interface_set_status(iface->cfg, &smap);
1804 ovsrec_interface_set_status(iface->cfg, NULL);
1807 smap_destroy(&smap);
1809 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1810 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1812 ovsrec_interface_set_duplex(iface->cfg,
1813 netdev_features_is_full_duplex(current)
1815 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1817 ovsrec_interface_set_duplex(iface->cfg, NULL);
1818 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1821 error = netdev_get_mtu(iface->netdev, &mtu);
1824 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1826 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1829 error = netdev_get_etheraddr(iface->netdev, mac);
1831 char mac_string[32];
1833 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1834 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1836 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1840 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1843 iface_refresh_cfm_stats(struct iface *iface)
1845 const struct ovsrec_interface *cfg = iface->cfg;
1846 struct ofproto_cfm_status status;
1848 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1849 iface->ofp_port, &status)) {
1850 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1851 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1852 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1853 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1854 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1856 const char *reasons[CFM_FAULT_N_REASONS];
1857 int64_t cfm_health = status.health;
1858 bool faulted = status.faults != 0;
1861 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1864 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1865 int reason = 1 << i;
1866 if (status.faults & reason) {
1867 reasons[j++] = cfm_fault_reason_to_str(reason);
1870 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1872 if (status.remote_opstate >= 0) {
1873 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1874 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1876 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1879 ovsrec_interface_set_cfm_remote_mpids(cfg,
1880 (const int64_t *)status.rmps,
1882 if (cfm_health >= 0) {
1883 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1885 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1891 iface_refresh_stats(struct iface *iface)
1893 #define IFACE_STATS \
1894 IFACE_STAT(rx_packets, "rx_packets") \
1895 IFACE_STAT(tx_packets, "tx_packets") \
1896 IFACE_STAT(rx_bytes, "rx_bytes") \
1897 IFACE_STAT(tx_bytes, "tx_bytes") \
1898 IFACE_STAT(rx_dropped, "rx_dropped") \
1899 IFACE_STAT(tx_dropped, "tx_dropped") \
1900 IFACE_STAT(rx_errors, "rx_errors") \
1901 IFACE_STAT(tx_errors, "tx_errors") \
1902 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1903 IFACE_STAT(rx_over_errors, "rx_over_err") \
1904 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1905 IFACE_STAT(collisions, "collisions")
1907 #define IFACE_STAT(MEMBER, NAME) + 1
1908 enum { N_IFACE_STATS = IFACE_STATS };
1910 int64_t values[N_IFACE_STATS];
1911 char *keys[N_IFACE_STATS];
1914 struct netdev_stats stats;
1916 if (iface_is_synthetic(iface)) {
1920 /* Intentionally ignore return value, since errors will set 'stats' to
1921 * all-1s, and we will deal with that correctly below. */
1922 netdev_get_stats(iface->netdev, &stats);
1924 /* Copy statistics into keys[] and values[]. */
1926 #define IFACE_STAT(MEMBER, NAME) \
1927 if (stats.MEMBER != UINT64_MAX) { \
1929 values[n] = stats.MEMBER; \
1934 ovs_assert(n <= N_IFACE_STATS);
1936 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1941 br_refresh_stp_status(struct bridge *br)
1943 struct smap smap = SMAP_INITIALIZER(&smap);
1944 struct ofproto *ofproto = br->ofproto;
1945 struct ofproto_stp_status status;
1947 if (ofproto_get_stp_status(ofproto, &status)) {
1951 if (!status.enabled) {
1952 ovsrec_bridge_set_status(br->cfg, NULL);
1956 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1957 STP_ID_ARGS(status.bridge_id));
1958 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1959 STP_ID_ARGS(status.designated_root));
1960 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1962 ovsrec_bridge_set_status(br->cfg, &smap);
1963 smap_destroy(&smap);
1967 port_refresh_stp_status(struct port *port)
1969 struct ofproto *ofproto = port->bridge->ofproto;
1970 struct iface *iface;
1971 struct ofproto_port_stp_status status;
1973 int64_t int_values[3];
1976 if (port_is_synthetic(port)) {
1980 /* STP doesn't currently support bonds. */
1981 if (!list_is_singleton(&port->ifaces)) {
1982 ovsrec_port_set_status(port->cfg, NULL);
1986 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1988 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1992 if (!status.enabled) {
1993 ovsrec_port_set_status(port->cfg, NULL);
1994 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1998 /* Set Status column. */
2000 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2001 smap_add(&smap, "stp_state", stp_state_name(status.state));
2002 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2003 smap_add(&smap, "stp_role", stp_role_name(status.role));
2004 ovsrec_port_set_status(port->cfg, &smap);
2005 smap_destroy(&smap);
2007 /* Set Statistics column. */
2008 keys[0] = "stp_tx_count";
2009 int_values[0] = status.tx_count;
2010 keys[1] = "stp_rx_count";
2011 int_values[1] = status.rx_count;
2012 keys[2] = "stp_error_count";
2013 int_values[2] = status.error_count;
2015 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2016 ARRAY_SIZE(int_values));
2020 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2022 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2026 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2028 bool enable = enable_system_stats(cfg);
2030 system_stats_enable(enable);
2032 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2037 run_system_stats(void)
2039 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2042 stats = system_stats_run();
2044 struct ovsdb_idl_txn *txn;
2045 struct ovsdb_datum datum;
2047 txn = ovsdb_idl_txn_create(idl);
2048 ovsdb_datum_from_smap(&datum, stats);
2049 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2051 ovsdb_idl_txn_commit(txn);
2052 ovsdb_idl_txn_destroy(txn);
2058 static inline const char *
2059 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2062 case OFPCR12_ROLE_EQUAL:
2064 case OFPCR12_ROLE_MASTER:
2066 case OFPCR12_ROLE_SLAVE:
2068 case OFPCR12_ROLE_NOCHANGE:
2070 return "*** INVALID ROLE ***";
2075 refresh_controller_status(void)
2079 const struct ovsrec_controller *cfg;
2083 /* Accumulate status for controllers on all bridges. */
2084 HMAP_FOR_EACH (br, node, &all_bridges) {
2085 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2088 /* Update each controller in the database with current status. */
2089 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2090 struct ofproto_controller_info *cinfo =
2091 shash_find_data(&info, cfg->target);
2094 struct smap smap = SMAP_INITIALIZER(&smap);
2095 const char **values = cinfo->pairs.values;
2096 const char **keys = cinfo->pairs.keys;
2099 for (i = 0; i < cinfo->pairs.n; i++) {
2100 smap_add(&smap, keys[i], values[i]);
2103 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2104 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2106 ovsrec_controller_set_status(cfg, &smap);
2107 smap_destroy(&smap);
2109 ovsrec_controller_set_is_connected(cfg, false);
2110 ovsrec_controller_set_role(cfg, NULL);
2111 ovsrec_controller_set_status(cfg, NULL);
2115 ofproto_free_ofproto_controller_info(&info);
2120 * Some information in the database must be kept as up-to-date as possible to
2121 * allow controllers to respond rapidly to network outages. We call these
2122 * statistics "instant" stats.
2124 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2125 * assuming that they've changed. The only means we have to determine whether
2126 * they have changed are:
2128 * - Try to commit changes to the database. If nothing changed, then
2129 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2132 * - instant_stats_run() is called late in the run loop, after anything that
2133 * might change any of the instant stats.
2135 * We use these two facts together to avoid waking the process up every
2136 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2139 /* Minimum interval between writing updates to the instant stats to the
2141 #define INSTANT_INTERVAL_MSEC 100
2143 /* Current instant stats database transaction, NULL if there is no ongoing
2145 static struct ovsdb_idl_txn *instant_txn;
2147 /* Next time (in msec on monotonic clock) at which we will update the instant
2149 static long long int instant_next_txn = LLONG_MIN;
2151 /* True if the run loop has run since we last saw that the instant stats were
2152 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2153 * to refresh the instant stats. */
2154 static bool instant_stats_could_have_changed;
2157 instant_stats_run(void)
2159 enum ovsdb_idl_txn_status status;
2161 instant_stats_could_have_changed = true;
2166 if (time_msec() < instant_next_txn) {
2169 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2171 instant_txn = ovsdb_idl_txn_create(idl);
2172 HMAP_FOR_EACH (br, node, &all_bridges) {
2173 struct iface *iface;
2176 br_refresh_stp_status(br);
2178 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2179 port_refresh_stp_status(port);
2182 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2183 enum netdev_flags flags;
2185 const char *link_state;
2186 int64_t link_resets;
2189 if (iface_is_synthetic(iface)) {
2193 current = ofproto_port_is_lacp_current(br->ofproto,
2197 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2199 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2202 error = netdev_get_flags(iface->netdev, &flags);
2204 const char *state = flags & NETDEV_UP ? "up" : "down";
2205 ovsrec_interface_set_admin_state(iface->cfg, state);
2207 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2210 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2211 ovsrec_interface_set_link_state(iface->cfg, link_state);
2213 link_resets = netdev_get_carrier_resets(iface->netdev);
2214 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2216 iface_refresh_cfm_stats(iface);
2219 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2221 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2222 smap_destroy(&smap);
2228 status = ovsdb_idl_txn_commit(instant_txn);
2229 if (status != TXN_INCOMPLETE) {
2230 ovsdb_idl_txn_destroy(instant_txn);
2233 if (status == TXN_UNCHANGED) {
2234 instant_stats_could_have_changed = false;
2239 instant_stats_wait(void)
2242 ovsdb_idl_txn_wait(instant_txn);
2243 } else if (instant_stats_could_have_changed) {
2244 poll_timer_wait_until(instant_next_txn);
2248 /* Performs periodic activity required by bridges that needs to be done with
2249 * the least possible latency.
2251 * It makes sense to call this function a couple of times per poll loop, to
2252 * provide a significant performance boost on some benchmarks with ofprotos
2253 * that use the ofproto-dpif implementation. */
2255 bridge_run_fast(void)
2262 ofproto_enumerate_types(&types);
2263 SSET_FOR_EACH (type, &types) {
2264 ofproto_type_run_fast(type);
2266 sset_destroy(&types);
2268 HMAP_FOR_EACH (br, node, &all_bridges) {
2269 ofproto_run_fast(br->ofproto);
2276 static struct ovsrec_open_vswitch null_cfg;
2277 const struct ovsrec_open_vswitch *cfg;
2278 struct ovsdb_idl_txn *reconf_txn = NULL;
2282 bool vlan_splinters_changed;
2285 ovsrec_open_vswitch_init(&null_cfg);
2287 /* (Re)configure if necessary. */
2288 if (!reconfiguring) {
2291 if (ovsdb_idl_is_lock_contended(idl)) {
2292 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2293 struct bridge *br, *next_br;
2295 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2296 "disabling this process (pid %ld) until it goes away",
2297 (long int) getpid());
2299 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2302 /* Since we will not be running system_stats_run() in this process
2303 * with the current situation of multiple ovs-vswitchd daemons,
2304 * disable system stats collection. */
2305 system_stats_enable(false);
2307 } else if (!ovsdb_idl_has_lock(idl)) {
2311 cfg = ovsrec_open_vswitch_first(idl);
2313 /* Initialize the ofproto library. This only needs to run once, but
2314 * it must be done after the configuration is set. If the
2315 * initialization has already occurred, bridge_init_ofproto()
2316 * returns immediately. */
2317 bridge_init_ofproto(cfg);
2319 /* Once the value of flow-restore-wait is false, we no longer should
2320 * check its value from the database. */
2321 if (cfg && ofproto_get_flow_restore_wait()) {
2322 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2323 "flow-restore-wait", false));
2326 /* Let each datapath type do the work that it needs to do. */
2328 ofproto_enumerate_types(&types);
2329 SSET_FOR_EACH (type, &types) {
2330 ofproto_type_run(type);
2332 sset_destroy(&types);
2334 /* Let each bridge do the work that it needs to do. */
2335 HMAP_FOR_EACH (br, node, &all_bridges) {
2336 ofproto_run(br->ofproto);
2339 /* Re-configure SSL. We do this on every trip through the main loop,
2340 * instead of just when the database changes, because the contents of the
2341 * key and certificate files can change without the database changing.
2343 * We do this before bridge_reconfigure() because that function might
2344 * initiate SSL connections and thus requires SSL to be configured. */
2345 if (cfg && cfg->ssl) {
2346 const struct ovsrec_ssl *ssl = cfg->ssl;
2348 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2349 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2352 if (!reconfiguring) {
2353 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2354 * usage has changed. */
2355 vlan_splinters_changed = false;
2356 if (vlan_splinters_enabled_anywhere) {
2357 HMAP_FOR_EACH (br, node, &all_bridges) {
2358 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2359 vlan_splinters_changed = true;
2365 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2366 idl_seqno = ovsdb_idl_get_seqno(idl);
2368 reconf_txn = ovsdb_idl_txn_create(idl);
2369 bridge_reconfigure(cfg);
2371 /* We still need to reconfigure to avoid dangling pointers to
2372 * now-destroyed ovsrec structures inside bridge data. */
2373 bridge_reconfigure(&null_cfg);
2378 if (reconfiguring) {
2380 reconf_txn = ovsdb_idl_txn_create(idl);
2383 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2384 reconfiguring = false;
2387 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2390 /* If we are completing our initial configuration for this run
2391 * of ovs-vswitchd, then keep the transaction around to monitor
2392 * it for completion. */
2393 if (!initial_config_done) {
2394 initial_config_done = true;
2395 daemonize_txn = reconf_txn;
2402 ovsdb_idl_txn_commit(reconf_txn);
2403 ovsdb_idl_txn_destroy(reconf_txn);
2407 if (daemonize_txn) {
2408 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2409 if (status != TXN_INCOMPLETE) {
2410 ovsdb_idl_txn_destroy(daemonize_txn);
2411 daemonize_txn = NULL;
2413 /* ovs-vswitchd has completed initialization, so allow the
2414 * process that forked us to exit successfully. */
2415 daemonize_complete();
2417 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2421 /* Refresh interface and mirror stats if necessary. */
2422 if (time_msec() >= iface_stats_timer) {
2424 struct ovsdb_idl_txn *txn;
2426 txn = ovsdb_idl_txn_create(idl);
2427 HMAP_FOR_EACH (br, node, &all_bridges) {
2431 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2432 struct iface *iface;
2434 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2435 iface_refresh_stats(iface);
2436 iface_refresh_status(iface);
2440 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2441 mirror_refresh_stats(m);
2445 refresh_controller_status();
2446 ovsdb_idl_txn_commit(txn);
2447 ovsdb_idl_txn_destroy(txn); /* XXX */
2450 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2454 instant_stats_run();
2463 ovsdb_idl_wait(idl);
2464 if (daemonize_txn) {
2465 ovsdb_idl_txn_wait(daemonize_txn);
2468 if (reconfiguring) {
2469 poll_immediate_wake();
2473 ofproto_enumerate_types(&types);
2474 SSET_FOR_EACH (type, &types) {
2475 ofproto_type_wait(type);
2477 sset_destroy(&types);
2479 if (!hmap_is_empty(&all_bridges)) {
2482 HMAP_FOR_EACH (br, node, &all_bridges) {
2483 ofproto_wait(br->ofproto);
2485 poll_timer_wait_until(iface_stats_timer);
2488 system_stats_wait();
2489 instant_stats_wait();
2492 /* Adds some memory usage statistics for bridges into 'usage', for use with
2493 * memory_report(). */
2495 bridge_get_memory_usage(struct simap *usage)
2499 HMAP_FOR_EACH (br, node, &all_bridges) {
2500 ofproto_get_memory_usage(br->ofproto, usage);
2504 /* QoS unixctl user interface functions. */
2506 struct qos_unixctl_show_cbdata {
2508 struct iface *iface;
2512 qos_unixctl_show_cb(unsigned int queue_id,
2513 const struct smap *details,
2516 struct qos_unixctl_show_cbdata *data = aux;
2517 struct ds *ds = data->ds;
2518 struct iface *iface = data->iface;
2519 struct netdev_queue_stats stats;
2520 struct smap_node *node;
2523 ds_put_cstr(ds, "\n");
2525 ds_put_format(ds, "Queue %u:\n", queue_id);
2527 ds_put_cstr(ds, "Default:\n");
2530 SMAP_FOR_EACH (node, details) {
2531 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2534 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2536 if (stats.tx_packets != UINT64_MAX) {
2537 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2540 if (stats.tx_bytes != UINT64_MAX) {
2541 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2544 if (stats.tx_errors != UINT64_MAX) {
2545 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2548 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2549 queue_id, strerror(error));
2554 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2555 const char *argv[], void *aux OVS_UNUSED)
2557 struct ds ds = DS_EMPTY_INITIALIZER;
2558 struct smap smap = SMAP_INITIALIZER(&smap);
2559 struct iface *iface;
2561 struct smap_node *node;
2562 struct qos_unixctl_show_cbdata data;
2565 iface = iface_find(argv[1]);
2567 unixctl_command_reply_error(conn, "no such interface");
2571 netdev_get_qos(iface->netdev, &type, &smap);
2573 if (*type != '\0') {
2574 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2576 SMAP_FOR_EACH (node, &smap) {
2577 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2582 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2585 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2587 unixctl_command_reply(conn, ds_cstr(&ds));
2589 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2590 unixctl_command_reply_error(conn, ds_cstr(&ds));
2593 smap_destroy(&smap);
2597 /* Bridge reconfiguration functions. */
2599 bridge_create(const struct ovsrec_bridge *br_cfg)
2603 ovs_assert(!bridge_lookup(br_cfg->name));
2604 br = xzalloc(sizeof *br);
2606 br->name = xstrdup(br_cfg->name);
2607 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2610 /* Derive the default Ethernet address from the bridge's UUID. This should
2611 * be unique and it will be stable between ovs-vswitchd runs. */
2612 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2613 eth_addr_mark_random(br->default_ea);
2615 hmap_init(&br->ports);
2616 hmap_init(&br->ifaces);
2617 hmap_init(&br->iface_by_name);
2618 hmap_init(&br->mirrors);
2620 hmap_init(&br->if_cfg_todo);
2621 list_init(&br->ofpp_garbage);
2623 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2627 bridge_destroy(struct bridge *br)
2630 struct mirror *mirror, *next_mirror;
2631 struct port *port, *next_port;
2632 struct if_cfg *if_cfg, *next_if_cfg;
2633 struct ofpp_garbage *garbage, *next_garbage;
2635 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2638 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2639 mirror_destroy(mirror);
2641 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2642 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2645 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2646 &br->ofpp_garbage) {
2647 list_remove(&garbage->list_node);
2651 hmap_remove(&all_bridges, &br->node);
2652 ofproto_destroy(br->ofproto);
2653 hmap_destroy(&br->ifaces);
2654 hmap_destroy(&br->ports);
2655 hmap_destroy(&br->iface_by_name);
2656 hmap_destroy(&br->mirrors);
2657 hmap_destroy(&br->if_cfg_todo);
2664 static struct bridge *
2665 bridge_lookup(const char *name)
2669 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2670 if (!strcmp(br->name, name)) {
2677 /* Handle requests for a listing of all flows known by the OpenFlow
2678 * stack, including those normally hidden. */
2680 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2681 const char *argv[], void *aux OVS_UNUSED)
2686 br = bridge_lookup(argv[1]);
2688 unixctl_command_reply_error(conn, "Unknown bridge");
2693 ofproto_get_all_flows(br->ofproto, &results);
2695 unixctl_command_reply(conn, ds_cstr(&results));
2696 ds_destroy(&results);
2699 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2700 * connections and reconnect. If BRIDGE is not specified, then all bridges
2701 * drop their controller connections and reconnect. */
2703 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2704 const char *argv[], void *aux OVS_UNUSED)
2708 br = bridge_lookup(argv[1]);
2710 unixctl_command_reply_error(conn, "Unknown bridge");
2713 ofproto_reconnect_controllers(br->ofproto);
2715 HMAP_FOR_EACH (br, node, &all_bridges) {
2716 ofproto_reconnect_controllers(br->ofproto);
2719 unixctl_command_reply(conn, NULL);
2723 bridge_get_controllers(const struct bridge *br,
2724 struct ovsrec_controller ***controllersp)
2726 struct ovsrec_controller **controllers;
2727 size_t n_controllers;
2729 controllers = br->cfg->controller;
2730 n_controllers = br->cfg->n_controller;
2732 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2738 *controllersp = controllers;
2740 return n_controllers;
2744 bridge_queue_if_cfg(struct bridge *br,
2745 const struct ovsrec_interface *cfg,
2746 const struct ovsrec_port *parent)
2748 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2751 if_cfg->parent = parent;
2752 if_cfg->ofport = iface_pick_ofport(cfg);
2753 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2754 hash_string(if_cfg->cfg->name, 0));
2757 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2758 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2759 * 'br' needs to complete its configuration. */
2761 bridge_add_del_ports(struct bridge *br,
2762 const unsigned long int *splinter_vlans)
2764 struct shash_node *port_node;
2765 struct port *port, *next;
2766 struct shash new_ports;
2769 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2771 /* Collect new ports. */
2772 shash_init(&new_ports);
2773 for (i = 0; i < br->cfg->n_ports; i++) {
2774 const char *name = br->cfg->ports[i]->name;
2775 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2776 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2780 if (bridge_get_controllers(br, NULL)
2781 && !shash_find(&new_ports, br->name)) {
2782 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2783 br->name, br->name);
2785 ovsrec_interface_init(&br->synth_local_iface);
2786 ovsrec_port_init(&br->synth_local_port);
2788 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2789 br->synth_local_port.n_interfaces = 1;
2790 br->synth_local_port.name = br->name;
2792 br->synth_local_iface.name = br->name;
2793 br->synth_local_iface.type = "internal";
2795 br->synth_local_ifacep = &br->synth_local_iface;
2797 shash_add(&new_ports, br->name, &br->synth_local_port);
2800 if (splinter_vlans) {
2801 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2804 /* Get rid of deleted ports.
2805 * Get rid of deleted interfaces on ports that still exist. */
2806 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2807 port->cfg = shash_find_data(&new_ports, port->name);
2811 port_del_ifaces(port);
2815 /* Update iface->cfg and iface->type in interfaces that still exist.
2816 * Add new interfaces to creation queue. */
2817 SHASH_FOR_EACH (port_node, &new_ports) {
2818 const struct ovsrec_port *port = port_node->data;
2821 for (i = 0; i < port->n_interfaces; i++) {
2822 const struct ovsrec_interface *cfg = port->interfaces[i];
2823 struct iface *iface = iface_lookup(br, cfg->name);
2824 const char *type = iface_get_type(cfg, br->cfg);
2829 } else if (!strcmp(type, "null")) {
2830 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2831 " may be removed in February 2013. Please email"
2832 " dev@openvswitch.org with concerns.",
2835 bridge_queue_if_cfg(br, cfg, port);
2840 shash_destroy(&new_ports);
2843 /* Initializes 'oc' appropriately as a management service controller for
2846 * The caller must free oc->target when it is no longer needed. */
2848 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2849 struct ofproto_controller *oc)
2851 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2852 oc->max_backoff = 0;
2853 oc->probe_interval = 60;
2854 oc->band = OFPROTO_OUT_OF_BAND;
2856 oc->burst_limit = 0;
2857 oc->enable_async_msgs = true;
2860 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2862 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2863 struct ofproto_controller *oc)
2867 oc->target = c->target;
2868 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2869 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2870 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2871 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2872 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2873 oc->burst_limit = (c->controller_burst_limit
2874 ? *c->controller_burst_limit : 0);
2875 oc->enable_async_msgs = (!c->enable_async_messages
2876 || *c->enable_async_messages);
2877 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2878 if (dscp < 0 || dscp > 63) {
2879 dscp = DSCP_DEFAULT;
2884 /* Configures the IP stack for 'br''s local interface properly according to the
2885 * configuration in 'c'. */
2887 bridge_configure_local_iface_netdev(struct bridge *br,
2888 struct ovsrec_controller *c)
2890 struct netdev *netdev;
2891 struct in_addr mask, gateway;
2893 struct iface *local_iface;
2896 /* If there's no local interface or no IP address, give up. */
2897 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2898 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2902 /* Bring up the local interface. */
2903 netdev = local_iface->netdev;
2904 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2906 /* Configure the IP address and netmask. */
2907 if (!c->local_netmask
2908 || !inet_aton(c->local_netmask, &mask)
2910 mask.s_addr = guess_netmask(ip.s_addr);
2912 if (!netdev_set_in4(netdev, ip, mask)) {
2913 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2914 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2917 /* Configure the default gateway. */
2918 if (c->local_gateway
2919 && inet_aton(c->local_gateway, &gateway)
2920 && gateway.s_addr) {
2921 if (!netdev_add_router(netdev, gateway)) {
2922 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2923 br->name, IP_ARGS(gateway.s_addr));
2928 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2929 * in either string are treated as equal to any number of slashes in the other,
2930 * e.g. "x///y" is equal to "x/y".
2932 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2933 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2934 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2935 * 'b' against a prefix of 'a'.
2938 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2940 const char *b_start = b;
2942 if (b - b_start >= b_stoplen) {
2944 } else if (*a != *b) {
2946 } else if (*a == '/') {
2947 a += strspn(a, "/");
2948 b += strspn(b, "/");
2949 } else if (*a == '\0') {
2959 bridge_configure_remotes(struct bridge *br,
2960 const struct sockaddr_in *managers, size_t n_managers)
2962 bool disable_in_band;
2964 struct ovsrec_controller **controllers;
2965 size_t n_controllers;
2967 enum ofproto_fail_mode fail_mode;
2969 struct ofproto_controller *ocs;
2973 /* Check if we should disable in-band control on this bridge. */
2974 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2977 /* Set OpenFlow queue ID for in-band control. */
2978 ofproto_set_in_band_queue(br->ofproto,
2979 smap_get_int(&br->cfg->other_config,
2980 "in-band-queue", -1));
2982 if (disable_in_band) {
2983 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2985 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2988 n_controllers = bridge_get_controllers(br, &controllers);
2990 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2993 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2994 for (i = 0; i < n_controllers; i++) {
2995 struct ovsrec_controller *c = controllers[i];
2997 if (!strncmp(c->target, "punix:", 6)
2998 || !strncmp(c->target, "unix:", 5)) {
2999 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3002 if (!strncmp(c->target, "unix:", 5)) {
3003 /* Connect to a listening socket */
3004 whitelist = xasprintf("unix:%s/", ovs_rundir());
3005 if (strchr(c->target, '/') &&
3006 !equal_pathnames(c->target, whitelist,
3007 strlen(whitelist))) {
3008 /* Absolute path specified, but not in ovs_rundir */
3009 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3010 "controller \"%s\" due to possibility for "
3011 "remote exploit. Instead, specify socket "
3012 "in whitelisted \"%s\" or connect to "
3013 "\"unix:%s/%s.mgmt\" (which is always "
3014 "available without special configuration).",
3015 br->name, c->target, whitelist,
3016 ovs_rundir(), br->name);
3021 whitelist = xasprintf("punix:%s/%s.controller",
3022 ovs_rundir(), br->name);
3023 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3024 /* Prevent remote ovsdb-server users from accessing
3025 * arbitrary Unix domain sockets and overwriting arbitrary
3027 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3028 "controller \"%s\" due to possibility of "
3029 "overwriting local files. Instead, specify "
3030 "whitelisted \"%s\" or connect to "
3031 "\"unix:%s/%s.mgmt\" (which is always "
3032 "available without special configuration).",
3033 br->name, c->target, whitelist,
3034 ovs_rundir(), br->name);
3043 bridge_configure_local_iface_netdev(br, c);
3044 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3045 if (disable_in_band) {
3046 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3051 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3052 bridge_get_allowed_versions(br));
3053 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3056 /* Set the fail-mode. */
3057 fail_mode = !br->cfg->fail_mode
3058 || !strcmp(br->cfg->fail_mode, "standalone")
3059 ? OFPROTO_FAIL_STANDALONE
3060 : OFPROTO_FAIL_SECURE;
3061 ofproto_set_fail_mode(br->ofproto, fail_mode);
3063 /* Configure OpenFlow controller connection snooping. */
3064 if (!ofproto_has_snoops(br->ofproto)) {
3068 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3069 ovs_rundir(), br->name));
3070 ofproto_set_snoops(br->ofproto, &snoops);
3071 sset_destroy(&snoops);
3076 bridge_configure_tables(struct bridge *br)
3078 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3082 n_tables = ofproto_get_n_tables(br->ofproto);
3084 for (i = 0; i < n_tables; i++) {
3085 struct ofproto_table_settings s;
3088 s.max_flows = UINT_MAX;
3092 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3093 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3096 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3097 s.max_flows = *cfg->flow_limit;
3099 if (cfg->overflow_policy
3100 && !strcmp(cfg->overflow_policy, "evict")) {
3103 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3104 for (k = 0; k < cfg->n_groups; k++) {
3105 const char *string = cfg->groups[k];
3108 msg = mf_parse_subfield__(&s.groups[k], &string);
3110 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3111 "'groups' (%s)", br->name, i, msg);
3113 } else if (*string) {
3114 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3115 "element '%s' contains trailing garbage",
3116 br->name, i, cfg->groups[k]);
3124 ofproto_configure_table(br->ofproto, i, &s);
3128 for (; j < br->cfg->n_flow_tables; j++) {
3129 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3130 "%"PRId64" not supported by this datapath", br->name,
3131 br->cfg->key_flow_tables[j]);
3136 bridge_configure_dp_desc(struct bridge *br)
3138 ofproto_set_dp_desc(br->ofproto,
3139 smap_get(&br->cfg->other_config, "dp-desc"));
3142 /* Port functions. */
3144 static struct port *
3145 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3149 port = xzalloc(sizeof *port);
3151 port->name = xstrdup(cfg->name);
3153 list_init(&port->ifaces);
3155 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3159 /* Deletes interfaces from 'port' that are no longer configured for it. */
3161 port_del_ifaces(struct port *port)
3163 struct iface *iface, *next;
3164 struct sset new_ifaces;
3167 /* Collect list of new interfaces. */
3168 sset_init(&new_ifaces);
3169 for (i = 0; i < port->cfg->n_interfaces; i++) {
3170 const char *name = port->cfg->interfaces[i]->name;
3171 const char *type = port->cfg->interfaces[i]->type;
3172 if (strcmp(type, "null")) {
3173 sset_add(&new_ifaces, name);
3177 /* Get rid of deleted interfaces. */
3178 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3179 if (!sset_contains(&new_ifaces, iface->name)) {
3180 iface_destroy(iface);
3184 sset_destroy(&new_ifaces);
3188 port_destroy(struct port *port)
3191 struct bridge *br = port->bridge;
3192 struct iface *iface, *next;
3195 ofproto_bundle_unregister(br->ofproto, port);
3198 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3199 iface_destroy(iface);
3202 hmap_remove(&br->ports, &port->hmap_node);
3208 static struct port *
3209 port_lookup(const struct bridge *br, const char *name)
3213 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3215 if (!strcmp(port->name, name)) {
3223 enable_lacp(struct port *port, bool *activep)
3225 if (!port->cfg->lacp) {
3226 /* XXX when LACP implementation has been sufficiently tested, enable by
3227 * default and make active on bonded ports. */
3229 } else if (!strcmp(port->cfg->lacp, "off")) {
3231 } else if (!strcmp(port->cfg->lacp, "active")) {
3234 } else if (!strcmp(port->cfg->lacp, "passive")) {
3238 VLOG_WARN("port %s: unknown LACP mode %s",
3239 port->name, port->cfg->lacp);
3244 static struct lacp_settings *
3245 port_configure_lacp(struct port *port, struct lacp_settings *s)
3247 const char *lacp_time, *system_id;
3250 if (!enable_lacp(port, &s->active)) {
3254 s->name = port->name;
3256 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3258 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3259 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3260 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3261 " address.", port->name, system_id);
3265 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3268 if (eth_addr_is_zero(s->id)) {
3269 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3273 /* Prefer bondable links if unspecified. */
3274 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3276 s->priority = (priority > 0 && priority <= UINT16_MAX
3278 : UINT16_MAX - !list_is_short(&port->ifaces));
3280 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3281 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3286 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3288 int priority, portid, key;
3290 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3291 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3293 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3295 if (portid <= 0 || portid > UINT16_MAX) {
3296 portid = ofp_to_u16(iface->ofp_port);
3299 if (priority <= 0 || priority > UINT16_MAX) {
3300 priority = UINT16_MAX;
3303 if (key < 0 || key > UINT16_MAX) {
3307 s->name = iface->name;
3309 s->priority = priority;
3314 port_configure_bond(struct port *port, struct bond_settings *s)
3316 const char *detect_s;
3317 struct iface *iface;
3318 int miimon_interval;
3320 s->name = port->name;
3322 if (port->cfg->bond_mode) {
3323 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3324 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3325 port->name, port->cfg->bond_mode,
3326 bond_mode_to_string(s->balance));
3329 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3331 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3332 * active-backup. At some point we should remove this warning. */
3333 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3334 " in previous versions, the default bond_mode was"
3335 " balance-slb", port->name,
3336 bond_mode_to_string(s->balance));
3338 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3339 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3340 "please use another bond type or disable flood_vlans",
3344 miimon_interval = smap_get_int(&port->cfg->other_config,
3345 "bond-miimon-interval", 0);
3346 if (miimon_interval <= 0) {
3347 miimon_interval = 200;
3350 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3351 if (!detect_s || !strcmp(detect_s, "carrier")) {
3352 miimon_interval = 0;
3353 } else if (strcmp(detect_s, "miimon")) {
3354 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3355 "defaulting to carrier", port->name, detect_s);
3356 miimon_interval = 0;
3359 s->up_delay = MAX(0, port->cfg->bond_updelay);
3360 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3361 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3362 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3363 "bond-rebalance-interval", 10000);
3364 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3365 s->rebalance_interval = 1000;
3368 s->fake_iface = port->cfg->bond_fake_iface;
3370 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3371 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3375 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3376 * instead of obtaining it from the database. */
3378 port_is_synthetic(const struct port *port)
3380 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3383 /* Interface functions. */
3386 iface_is_internal(const struct ovsrec_interface *iface,
3387 const struct ovsrec_bridge *br)
3389 /* The local port and "internal" ports are always "internal". */
3390 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3393 /* Returns the correct network device type for interface 'iface' in bridge
3396 iface_get_type(const struct ovsrec_interface *iface,
3397 const struct ovsrec_bridge *br)
3401 /* The local port always has type "internal". Other ports take
3402 * their type from the database and default to "system" if none is
3404 if (iface_is_internal(iface, br)) {
3407 type = iface->type[0] ? iface->type : "system";
3410 return ofproto_port_open_type(br->datapath_type, type);
3414 iface_destroy(struct iface *iface)
3417 struct port *port = iface->port;
3418 struct bridge *br = port->bridge;
3420 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3421 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3424 if (iface->ofp_port != OFPP_NONE) {
3425 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3428 list_remove(&iface->port_elem);
3429 hmap_remove(&br->iface_by_name, &iface->name_node);
3431 netdev_close(iface->netdev);
3438 static struct iface *
3439 iface_lookup(const struct bridge *br, const char *name)
3441 struct iface *iface;
3443 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3444 &br->iface_by_name) {
3445 if (!strcmp(iface->name, name)) {
3453 static struct iface *
3454 iface_find(const char *name)
3456 const struct bridge *br;
3458 HMAP_FOR_EACH (br, node, &all_bridges) {
3459 struct iface *iface = iface_lookup(br, name);
3468 static struct if_cfg *
3469 if_cfg_lookup(const struct bridge *br, const char *name)
3471 struct if_cfg *if_cfg;
3473 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3475 if (!strcmp(if_cfg->cfg->name, name)) {
3483 static struct iface *
3484 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3486 struct iface *iface;
3488 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3489 hash_int(ofp_to_u16(ofp_port), 0),
3491 if (iface->ofp_port == ofp_port) {
3498 /* Set Ethernet address of 'iface', if one is specified in the configuration
3501 iface_set_mac(struct iface *iface)
3503 uint8_t ea[ETH_ADDR_LEN];
3505 if (!strcmp(iface->type, "internal")
3506 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3507 if (iface->ofp_port == OFPP_LOCAL) {
3508 VLOG_ERR("interface %s: ignoring mac in Interface record "
3509 "(use Bridge record to set local port's mac)",
3511 } else if (eth_addr_is_multicast(ea)) {
3512 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3515 int error = netdev_set_etheraddr(iface->netdev, ea);
3517 VLOG_ERR("interface %s: setting MAC failed (%s)",
3518 iface->name, strerror(error));
3524 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3526 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3529 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3530 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3531 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3535 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3536 * sets the "ofport" field to -1.
3538 * This is appropriate when 'if_cfg''s interface cannot be created or is
3539 * otherwise invalid. */
3541 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3543 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3544 ovsrec_interface_set_status(if_cfg, NULL);
3545 ovsrec_interface_set_admin_state(if_cfg, NULL);
3546 ovsrec_interface_set_duplex(if_cfg, NULL);
3547 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3548 ovsrec_interface_set_link_state(if_cfg, NULL);
3549 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3550 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3551 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3552 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3553 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3554 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3555 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3559 struct iface_delete_queues_cbdata {
3560 struct netdev *netdev;
3561 const struct ovsdb_datum *queues;
3565 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3567 union ovsdb_atom atom;
3569 atom.integer = target;
3570 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3574 iface_delete_queues(unsigned int queue_id,
3575 const struct smap *details OVS_UNUSED, void *cbdata_)
3577 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3579 if (!queue_ids_include(cbdata->queues, queue_id)) {
3580 netdev_delete_queue(cbdata->netdev, queue_id);
3585 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3587 struct ofpbuf queues_buf;
3589 ofpbuf_init(&queues_buf, 0);
3591 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3592 netdev_set_qos(iface->netdev, NULL, NULL);
3594 struct iface_delete_queues_cbdata cbdata;
3598 /* Configure top-level Qos for 'iface'. */
3599 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3601 /* Deconfigure queues that were deleted. */
3602 cbdata.netdev = iface->netdev;
3603 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3605 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3607 /* Configure queues for 'iface'. */
3609 for (i = 0; i < qos->n_queues; i++) {
3610 const struct ovsrec_queue *queue = qos->value_queues[i];
3611 unsigned int queue_id = qos->key_queues[i];
3613 if (queue_id == 0) {
3617 if (queue->n_dscp == 1) {
3618 struct ofproto_port_queue *port_queue;
3620 port_queue = ofpbuf_put_uninit(&queues_buf,
3621 sizeof *port_queue);
3622 port_queue->queue = queue_id;
3623 port_queue->dscp = queue->dscp[0];
3626 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3629 struct smap details;
3631 smap_init(&details);
3632 netdev_set_queue(iface->netdev, 0, &details);
3633 smap_destroy(&details);
3637 if (iface->ofp_port != OFPP_NONE) {
3638 const struct ofproto_port_queue *port_queues = queues_buf.data;
3639 size_t n_queues = queues_buf.size / sizeof *port_queues;
3641 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3642 port_queues, n_queues);
3645 netdev_set_policing(iface->netdev,
3646 iface->cfg->ingress_policing_rate,
3647 iface->cfg->ingress_policing_burst);
3649 ofpbuf_uninit(&queues_buf);
3653 iface_configure_cfm(struct iface *iface)
3655 const struct ovsrec_interface *cfg = iface->cfg;
3656 const char *opstate_str;
3657 const char *cfm_ccm_vlan;
3658 struct cfm_settings s;
3659 struct smap netdev_args;
3661 if (!cfg->n_cfm_mpid) {
3662 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3666 s.check_tnl_key = false;
3667 smap_init(&netdev_args);
3668 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3669 const char *key = smap_get(&netdev_args, "key");
3670 const char *in_key = smap_get(&netdev_args, "in_key");
3672 s.check_tnl_key = (key && !strcmp(key, "flow"))
3673 || (in_key && !strcmp(in_key, "flow"));
3675 smap_destroy(&netdev_args);
3677 s.mpid = *cfg->cfm_mpid;
3678 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3679 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3680 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3682 if (s.interval <= 0) {
3686 if (!cfm_ccm_vlan) {
3688 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3689 s.ccm_vlan = CFM_RANDOM_VLAN;
3691 s.ccm_vlan = atoi(cfm_ccm_vlan);
3692 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3697 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3699 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3701 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3702 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3704 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3707 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3708 * instead of obtaining it from the database. */
3710 iface_is_synthetic(const struct iface *iface)
3712 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3716 iface_pick_ofport(const struct ovsrec_interface *cfg)
3718 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3720 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3725 /* Port mirroring. */
3727 static struct mirror *
3728 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3732 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3733 if (uuid_equals(uuid, &m->uuid)) {
3741 bridge_configure_mirrors(struct bridge *br)
3743 const struct ovsdb_datum *mc;
3744 unsigned long *flood_vlans;
3745 struct mirror *m, *next;
3748 /* Get rid of deleted mirrors. */
3749 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3750 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3751 union ovsdb_atom atom;
3753 atom.uuid = m->uuid;
3754 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3759 /* Add new mirrors and reconfigure existing ones. */
3760 for (i = 0; i < br->cfg->n_mirrors; i++) {
3761 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3762 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3764 m = mirror_create(br, cfg);
3767 if (!mirror_configure(m)) {
3772 /* Update flooded vlans (for RSPAN). */
3773 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3774 br->cfg->n_flood_vlans);
3775 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3776 bitmap_free(flood_vlans);
3779 static struct mirror *
3780 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3784 m = xzalloc(sizeof *m);
3785 m->uuid = cfg->header_.uuid;
3786 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3788 m->name = xstrdup(cfg->name);
3794 mirror_destroy(struct mirror *m)
3797 struct bridge *br = m->bridge;
3800 ofproto_mirror_unregister(br->ofproto, m);
3803 hmap_remove(&br->mirrors, &m->hmap_node);
3810 mirror_collect_ports(struct mirror *m,
3811 struct ovsrec_port **in_ports, int n_in_ports,
3812 void ***out_portsp, size_t *n_out_portsp)
3814 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3815 size_t n_out_ports = 0;
3818 for (i = 0; i < n_in_ports; i++) {
3819 const char *name = in_ports[i]->name;
3820 struct port *port = port_lookup(m->bridge, name);
3822 out_ports[n_out_ports++] = port;
3824 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3825 "port %s", m->bridge->name, m->name, name);
3828 *out_portsp = out_ports;
3829 *n_out_portsp = n_out_ports;
3833 mirror_configure(struct mirror *m)
3835 const struct ovsrec_mirror *cfg = m->cfg;
3836 struct ofproto_mirror_settings s;
3839 if (strcmp(cfg->name, m->name)) {
3841 m->name = xstrdup(cfg->name);
3845 /* Get output port or VLAN. */
3846 if (cfg->output_port) {
3847 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3848 if (!s.out_bundle) {
3849 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3850 m->bridge->name, m->name);
3853 s.out_vlan = UINT16_MAX;
3855 if (cfg->output_vlan) {
3856 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3857 "output vlan; ignoring output vlan",
3858 m->bridge->name, m->name);
3860 } else if (cfg->output_vlan) {
3861 /* The database should prevent invalid VLAN values. */
3862 s.out_bundle = NULL;
3863 s.out_vlan = *cfg->output_vlan;
3865 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3866 m->bridge->name, m->name);
3870 /* Get port selection. */
3871 if (cfg->select_all) {
3872 size_t n_ports = hmap_count(&m->bridge->ports);
3873 void **ports = xmalloc(n_ports * sizeof *ports);
3878 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3888 /* Get ports, dropping ports that don't exist.
3889 * The IDL ensures that there are no duplicates. */
3890 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3891 &s.srcs, &s.n_srcs);
3892 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3893 &s.dsts, &s.n_dsts);
3896 /* Get VLAN selection. */
3897 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3900 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3903 if (s.srcs != s.dsts) {
3912 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3914 * This is deprecated. It is only for compatibility with broken device drivers
3915 * in old versions of Linux that do not properly support VLANs when VLAN
3916 * devices are not used. When broken device drivers are no longer in
3917 * widespread use, we will delete these interfaces. */
3919 static struct ovsrec_port **recs;
3920 static size_t n_recs, allocated_recs;
3922 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3923 * splinters are reconfigured. */
3925 register_rec(struct ovsrec_port *rec)
3927 if (n_recs >= allocated_recs) {
3928 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3930 recs[n_recs++] = rec;
3933 /* Frees all of the ports registered with register_reg(). */
3935 free_registered_recs(void)
3939 for (i = 0; i < n_recs; i++) {
3940 struct ovsrec_port *port = recs[i];
3943 for (j = 0; j < port->n_interfaces; j++) {
3944 struct ovsrec_interface *iface = port->interfaces[j];
3949 smap_destroy(&port->other_config);
3950 free(port->interfaces);
3958 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3961 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3963 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3967 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3970 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3971 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3972 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3975 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3976 * use, returns NULL.
3978 * Updates 'vlan_splinters_enabled_anywhere'. */
3979 static unsigned long int *
3980 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3982 unsigned long int *splinter_vlans;
3983 struct sset splinter_ifaces;
3984 const char *real_dev_name;
3985 struct shash *real_devs;
3986 struct shash_node *node;
3990 /* Free space allocated for synthesized ports and interfaces, since we're
3991 * in the process of reconstructing all of them. */
3992 free_registered_recs();
3994 splinter_vlans = bitmap_allocate(4096);
3995 sset_init(&splinter_ifaces);
3996 vlan_splinters_enabled_anywhere = false;
3997 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3998 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4001 for (j = 0; j < br_cfg->n_ports; j++) {
4002 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4005 for (k = 0; k < port_cfg->n_interfaces; k++) {
4006 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4008 if (vlan_splinters_is_enabled(iface_cfg)) {
4009 vlan_splinters_enabled_anywhere = true;
4010 sset_add(&splinter_ifaces, iface_cfg->name);
4011 vlan_bitmap_from_array__(port_cfg->trunks,
4017 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4018 bitmap_set1(splinter_vlans, *port_cfg->tag);
4023 if (!vlan_splinters_enabled_anywhere) {
4024 free(splinter_vlans);
4025 sset_destroy(&splinter_ifaces);
4029 HMAP_FOR_EACH (br, node, &all_bridges) {
4031 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4035 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4036 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4037 * device to be created for it. */
4038 bitmap_set0(splinter_vlans, 0);
4039 bitmap_set0(splinter_vlans, 4095);
4041 /* Delete all VLAN devices that we don't need. */
4043 real_devs = vlandev_get_real_devs();
4044 SHASH_FOR_EACH (node, real_devs) {
4045 const struct vlan_real_dev *real_dev = node->data;
4046 const struct vlan_dev *vlan_dev;
4047 bool real_dev_has_splinters;
4049 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4051 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4052 if (!real_dev_has_splinters
4053 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4054 struct netdev *netdev;
4056 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4057 if (!netdev_get_in4(netdev, NULL, NULL) ||
4058 !netdev_get_in6(netdev, NULL)) {
4059 vlandev_del(vlan_dev->name);
4061 /* It has an IP address configured, so we don't own
4062 * it. Don't delete it. */
4064 netdev_close(netdev);
4071 /* Add all VLAN devices that we need. */
4072 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4075 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4076 if (!vlandev_get_name(real_dev_name, vid)) {
4077 vlandev_add(real_dev_name, vid);
4084 sset_destroy(&splinter_ifaces);
4086 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4087 free(splinter_vlans);
4090 return splinter_vlans;
4093 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4096 configure_splinter_port(struct port *port)
4098 struct ofproto *ofproto = port->bridge->ofproto;
4099 ofp_port_t realdev_ofp_port;
4100 const char *realdev_name;
4101 struct iface *vlandev, *realdev;
4103 ofproto_bundle_unregister(port->bridge->ofproto, port);
4105 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4108 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4109 realdev = iface_lookup(port->bridge, realdev_name);
4110 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4112 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4116 static struct ovsrec_port *
4117 synthesize_splinter_port(const char *real_dev_name,
4118 const char *vlan_dev_name, int vid)
4120 struct ovsrec_interface *iface;
4121 struct ovsrec_port *port;
4123 iface = xmalloc(sizeof *iface);
4124 ovsrec_interface_init(iface);
4125 iface->name = xstrdup(vlan_dev_name);
4126 iface->type = "system";
4128 port = xmalloc(sizeof *port);
4129 ovsrec_port_init(port);
4130 port->interfaces = xmemdup(&iface, sizeof iface);
4131 port->n_interfaces = 1;
4132 port->name = xstrdup(vlan_dev_name);
4133 port->vlan_mode = "splinter";
4134 port->tag = xmalloc(sizeof *port->tag);
4137 smap_add(&port->other_config, "realdev", real_dev_name);
4143 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4144 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4145 * 1-bit in the 'splinter_vlans' bitmap. */
4147 add_vlan_splinter_ports(struct bridge *br,
4148 const unsigned long int *splinter_vlans,
4149 struct shash *ports)
4153 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4154 * we're modifying 'ports'. */
4155 for (i = 0; i < br->cfg->n_ports; i++) {
4156 const char *name = br->cfg->ports[i]->name;
4157 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4160 for (j = 0; j < port_cfg->n_interfaces; j++) {
4161 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4163 if (vlan_splinters_is_enabled(iface_cfg)) {
4164 const char *real_dev_name;
4167 real_dev_name = iface_cfg->name;
4168 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4169 const char *vlan_dev_name;
4171 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4173 && !shash_find(ports, vlan_dev_name)) {
4174 shash_add(ports, vlan_dev_name,
4175 synthesize_splinter_port(
4176 real_dev_name, vlan_dev_name, vid));
4185 mirror_refresh_stats(struct mirror *m)
4187 struct ofproto *ofproto = m->bridge->ofproto;
4188 uint64_t tx_packets, tx_bytes;
4191 size_t stat_cnt = 0;
4193 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4194 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4198 if (tx_packets != UINT64_MAX) {
4199 keys[stat_cnt] = "tx_packets";
4200 values[stat_cnt] = tx_packets;
4203 if (tx_bytes != UINT64_MAX) {
4204 keys[stat_cnt] = "tx_bytes";
4205 values[stat_cnt] = tx_bytes;
4209 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);