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.
27 #include "dynamic-string.h"
34 #include "mac-learning.h"
35 #include "meta-flow.h"
37 #include "ofp-print.h"
40 #include "ofproto/ofproto.h"
41 #include "poll-loop.h"
45 #include "socket-util.h"
47 #include "stream-ssl.h"
49 #include "system-stats.h"
54 #include "lib/vswitch-idl.h"
55 #include "xenserver.h"
57 #include "sflow_api.h"
58 #include "vlan-bitmap.h"
60 VLOG_DEFINE_THIS_MODULE(bridge);
62 COVERAGE_DEFINE(bridge_reconfigure);
64 /* Configuration of an uninstantiated iface. */
66 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
67 const struct ovsrec_interface *cfg; /* Interface record. */
68 const struct ovsrec_port *parent; /* Parent port record. */
69 int64_t ofport; /* Requested OpenFlow port number. */
72 /* OpenFlow port slated for removal from ofproto. */
74 struct list list_node; /* Node in bridge's ofpp_garbage. */
75 uint16_t ofp_port; /* Port to be deleted. */
79 /* These members are always valid. */
80 struct list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct port *port; /* Containing port. */
83 char *name; /* Host network device name. */
85 /* These members are valid only after bridge_reconfigure() causes them to
87 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
88 int ofp_port; /* OpenFlow port number, -1 if unknown. */
89 struct netdev *netdev; /* Network device. */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
131 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
132 Indexed on 'cfg->name'. */
134 /* Port mirroring. */
135 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
137 /* Synthetic local port if necessary. */
138 struct ovsrec_port synth_local_port;
139 struct ovsrec_interface synth_local_iface;
140 struct ovsrec_interface *synth_local_ifacep;
143 /* All bridges, indexed by name. */
144 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
146 /* OVSDB IDL used to obtain configuration. */
147 static struct ovsdb_idl *idl;
149 /* We want to complete daemonization, fully detaching from our parent process,
150 * only after we have completed our initial configuration, committed our state
151 * to the database, and received confirmation back from the database server
152 * that it applied the commit. This allows our parent process to know that,
153 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
154 * to have already been filled in. (It is only "very likely" rather than
155 * certain because there is always a slim possibility that the transaction will
156 * fail or that some other client has added new bridges, ports, etc. while
157 * ovs-vswitchd was configuring using an old configuration.)
159 * We only need to do this once for our initial configuration at startup, so
160 * 'initial_config_done' tracks whether we've already done it. While we are
161 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
162 * itself and is otherwise NULL. */
163 static bool initial_config_done;
164 static struct ovsdb_idl_txn *daemonize_txn;
166 /* Most recently processed IDL sequence number. */
167 static unsigned int idl_seqno;
169 /* Each time this timer expires, the bridge fetches interface and mirror
170 * statistics and pushes them into the database. */
171 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
172 static long long int iface_stats_timer = LLONG_MIN;
174 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
175 * expensive. This can cause bridge_reconfigure() to take a long time during
176 * which no other work can be done. To deal with this problem, we limit port
177 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
178 * call to bridge_reconfigure(). If there is more work to do after the limit
179 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
180 * This allows the rest of the code to catch up on important things like
181 * forwarding packets. */
182 #define OFP_PORT_ACTION_WINDOW 10
183 static bool reconfiguring = false;
185 static void add_del_bridges(const struct ovsrec_open_vswitch *);
186 static void bridge_update_ofprotos(void);
187 static void bridge_create(const struct ovsrec_bridge *);
188 static void bridge_destroy(struct bridge *);
189 static struct bridge *bridge_lookup(const char *name);
190 static unixctl_cb_func bridge_unixctl_dump_flows;
191 static unixctl_cb_func bridge_unixctl_reconnect;
192 static size_t bridge_get_controllers(const struct bridge *br,
193 struct ovsrec_controller ***controllersp);
194 static void bridge_add_del_ports(struct bridge *,
195 const unsigned long int *splinter_vlans);
196 static void bridge_refresh_ofp_port(struct bridge *);
197 static void bridge_configure_datapath_id(struct bridge *);
198 static void bridge_configure_flow_eviction_threshold(struct bridge *);
199 static void bridge_configure_netflow(struct bridge *);
200 static void bridge_configure_forward_bpdu(struct bridge *);
201 static void bridge_configure_mac_table(struct bridge *);
202 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
203 static void bridge_configure_ipfix(struct bridge *);
204 static void bridge_configure_stp(struct bridge *);
205 static void bridge_configure_tables(struct bridge *);
206 static void bridge_configure_dp_desc(struct bridge *);
207 static void bridge_configure_remotes(struct bridge *,
208 const struct sockaddr_in *managers,
210 static void bridge_pick_local_hw_addr(struct bridge *,
211 uint8_t ea[ETH_ADDR_LEN],
212 struct iface **hw_addr_iface);
213 static uint64_t bridge_pick_datapath_id(struct bridge *,
214 const uint8_t bridge_ea[ETH_ADDR_LEN],
215 struct iface *hw_addr_iface);
216 static void bridge_queue_if_cfg(struct bridge *,
217 const struct ovsrec_interface *,
218 const struct ovsrec_port *);
219 static uint64_t dpid_from_hash(const void *, size_t nbytes);
220 static bool bridge_has_bond_fake_iface(const struct bridge *,
222 static bool port_is_bond_fake_iface(const struct port *);
224 static unixctl_cb_func qos_unixctl_show;
226 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
227 static void port_del_ifaces(struct port *);
228 static void port_destroy(struct port *);
229 static struct port *port_lookup(const struct bridge *, const char *name);
230 static void port_configure(struct port *);
231 static struct lacp_settings *port_configure_lacp(struct port *,
232 struct lacp_settings *);
233 static void port_configure_bond(struct port *, struct bond_settings *);
234 static bool port_is_synthetic(const struct port *);
236 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
237 static void run_system_stats(void);
239 static void bridge_configure_mirrors(struct bridge *);
240 static struct mirror *mirror_create(struct bridge *,
241 const struct ovsrec_mirror *);
242 static void mirror_destroy(struct mirror *);
243 static bool mirror_configure(struct mirror *);
244 static void mirror_refresh_stats(struct mirror *);
246 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
247 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
248 static bool iface_is_internal(const struct ovsrec_interface *iface,
249 const struct ovsrec_bridge *br);
250 static const char *iface_get_type(const struct ovsrec_interface *,
251 const struct ovsrec_bridge *);
252 static void iface_destroy(struct iface *);
253 static struct iface *iface_lookup(const struct bridge *, const char *name);
254 static struct iface *iface_find(const char *name);
255 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
256 static struct iface *iface_from_ofp_port(const struct bridge *,
258 static void iface_set_mac(struct iface *);
259 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
260 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
261 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
262 static void iface_configure_cfm(struct iface *);
263 static void iface_refresh_cfm_stats(struct iface *);
264 static void iface_refresh_stats(struct iface *);
265 static void iface_refresh_status(struct iface *);
266 static bool iface_is_synthetic(const struct iface *);
267 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
269 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
271 * This is deprecated. It is only for compatibility with broken device drivers
272 * in old versions of Linux that do not properly support VLANs when VLAN
273 * devices are not used. When broken device drivers are no longer in
274 * widespread use, we will delete these interfaces. */
276 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
277 static bool vlan_splinters_enabled_anywhere;
279 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
280 static unsigned long int *collect_splinter_vlans(
281 const struct ovsrec_open_vswitch *);
282 static void configure_splinter_port(struct port *);
283 static void add_vlan_splinter_ports(struct bridge *,
284 const unsigned long int *splinter_vlans,
285 struct shash *ports);
288 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
290 struct shash iface_hints;
291 static bool initialized = false;
298 shash_init(&iface_hints);
301 for (i = 0; i < cfg->n_bridges; i++) {
302 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
305 for (j = 0; j < br_cfg->n_ports; j++) {
306 struct ovsrec_port *port_cfg = br_cfg->ports[j];
309 for (k = 0; k < port_cfg->n_interfaces; k++) {
310 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
311 struct iface_hint *iface_hint;
313 iface_hint = xmalloc(sizeof *iface_hint);
314 iface_hint->br_name = br_cfg->name;
315 iface_hint->br_type = br_cfg->datapath_type;
316 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
318 shash_add(&iface_hints, if_cfg->name, iface_hint);
324 ofproto_init(&iface_hints);
326 shash_destroy_free_data(&iface_hints);
330 /* Public functions. */
332 /* Initializes the bridge module, configuring it to obtain its configuration
333 * from an OVSDB server accessed over 'remote', which should be a string in a
334 * form acceptable to ovsdb_idl_create(). */
336 bridge_init(const char *remote)
338 /* Create connection to database. */
339 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
340 idl_seqno = ovsdb_idl_get_seqno(idl);
341 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
342 ovsdb_idl_verify_write_only(idl);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
345 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
353 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
354 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
357 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
358 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
359 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
377 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
382 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
384 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
386 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
389 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
391 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
402 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
404 /* Register unixctl commands. */
405 unixctl_command_register("qos/show", "interface", 1, 1,
406 qos_unixctl_show, NULL);
407 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
408 bridge_unixctl_dump_flows, NULL);
409 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
410 bridge_unixctl_reconnect, NULL);
420 struct bridge *br, *next_br;
422 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
425 ovsdb_idl_destroy(idl);
428 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
429 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
430 * responsible for freeing '*managersp' (with free()).
432 * You may be asking yourself "why does ovs-vswitchd care?", because
433 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
434 * should not be and in fact is not directly involved in that. But
435 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
436 * it has to tell in-band control where the managers are to enable that.
437 * (Thus, only managers connected in-band are collected.)
440 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
441 struct sockaddr_in **managersp, size_t *n_managersp)
443 struct sockaddr_in *managers = NULL;
444 size_t n_managers = 0;
448 /* Collect all of the potential targets from the "targets" columns of the
449 * rows pointed to by "manager_options", excluding any that are
452 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
453 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
455 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
456 sset_find_and_delete(&targets, m->target);
458 sset_add(&targets, m->target);
462 /* Now extract the targets' IP addresses. */
463 if (!sset_is_empty(&targets)) {
466 managers = xmalloc(sset_count(&targets) * sizeof *managers);
467 SSET_FOR_EACH (target, &targets) {
468 struct sockaddr_in *sin = &managers[n_managers];
470 if (stream_parse_target_with_default_ports(target,
478 sset_destroy(&targets);
480 *managersp = managers;
481 *n_managersp = n_managers;
485 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
487 unsigned long int *splinter_vlans;
490 COVERAGE_INC(bridge_reconfigure);
492 ovs_assert(!reconfiguring);
493 reconfiguring = true;
495 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
496 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
497 * configuration otherwise.
499 * This is mostly an update to bridge data structures. Nothing is pushed
500 * down to ofproto or lower layers. */
501 add_del_bridges(ovs_cfg);
502 splinter_vlans = collect_splinter_vlans(ovs_cfg);
503 HMAP_FOR_EACH (br, node, &all_bridges) {
504 bridge_add_del_ports(br, splinter_vlans);
506 free(splinter_vlans);
508 /* Delete datapaths that are no longer configured, and create ones which
509 * don't exist but should. */
510 bridge_update_ofprotos();
512 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
513 HMAP_FOR_EACH (br, node, &all_bridges) {
514 bridge_refresh_ofp_port(br);
517 /* Clear database records for "if_cfg"s which haven't been instantiated. */
518 HMAP_FOR_EACH (br, node, &all_bridges) {
519 struct if_cfg *if_cfg;
521 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
522 iface_clear_db_record(if_cfg->cfg);
526 reconfigure_system_stats(ovs_cfg);
530 bridge_reconfigure_ofp(void)
532 long long int deadline;
536 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
538 /* The kernel will reject any attempt to add a given port to a datapath if
539 * that port already belongs to a different datapath, so we must do all
540 * port deletions before any port additions. */
541 HMAP_FOR_EACH (br, node, &all_bridges) {
542 struct ofpp_garbage *garbage, *next;
544 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
545 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
546 * internal datastructures may not be consistent. Eventually, when
547 * port additions and deletions are cheaper, these calls should be
550 ofproto_port_del(br->ofproto, garbage->ofp_port);
551 list_remove(&garbage->list_node);
555 if (time_msec() >= deadline) {
562 HMAP_FOR_EACH (br, node, &all_bridges) {
563 struct if_cfg *if_cfg, *next;
565 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
566 iface_create(br, if_cfg, -1);
568 if (time_msec() >= deadline) {
578 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
580 struct sockaddr_in *managers;
581 int sflow_bridge_number;
586 ovs_assert(reconfiguring);
587 done = bridge_reconfigure_ofp();
589 /* Complete the configuration. */
590 sflow_bridge_number = 0;
591 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
592 HMAP_FOR_EACH (br, node, &all_bridges) {
595 /* We need the datapath ID early to allow LACP ports to use it as the
596 * default system ID. */
597 bridge_configure_datapath_id(br);
599 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
602 port_configure(port);
604 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
605 iface_configure_cfm(iface);
606 iface_configure_qos(iface, port->cfg->qos);
607 iface_set_mac(iface);
610 bridge_configure_mirrors(br);
611 bridge_configure_flow_eviction_threshold(br);
612 bridge_configure_forward_bpdu(br);
613 bridge_configure_mac_table(br);
614 bridge_configure_remotes(br, managers, n_managers);
615 bridge_configure_netflow(br);
616 bridge_configure_sflow(br, &sflow_bridge_number);
617 bridge_configure_ipfix(br);
618 bridge_configure_stp(br);
619 bridge_configure_tables(br);
620 bridge_configure_dp_desc(br);
627 /* Delete ofprotos which aren't configured or have the wrong type. Create
628 * ofprotos which don't exist but need to. */
630 bridge_update_ofprotos(void)
632 struct bridge *br, *next;
637 /* Delete ofprotos with no bridge or with the wrong type. */
640 ofproto_enumerate_types(&types);
641 SSET_FOR_EACH (type, &types) {
644 ofproto_enumerate_names(type, &names);
645 SSET_FOR_EACH (name, &names) {
646 br = bridge_lookup(name);
647 if (!br || strcmp(type, br->type)) {
648 ofproto_delete(name, type);
652 sset_destroy(&names);
653 sset_destroy(&types);
655 /* Add ofprotos for bridges which don't have one yet. */
656 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
664 /* Remove ports from any datapath with the same name as 'br'. If we
665 * don't do this, creating 'br''s ofproto will fail because a port with
666 * the same name as its local port already exists. */
667 HMAP_FOR_EACH (br2, node, &all_bridges) {
668 struct ofproto_port ofproto_port;
674 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
676 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
678 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
681 ofproto_port_destroy(&ofproto_port);
685 error = ofproto_create(br->name, br->type, &br->ofproto);
687 VLOG_ERR("failed to create bridge %s: %s", br->name,
695 port_configure(struct port *port)
697 const struct ovsrec_port *cfg = port->cfg;
698 struct bond_settings bond_settings;
699 struct lacp_settings lacp_settings;
700 struct ofproto_bundle_settings s;
703 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
704 configure_splinter_port(port);
713 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
714 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
715 s.slaves[s.n_slaves++] = iface->ofp_port;
720 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
724 /* Get VLAN trunks. */
727 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
731 if (cfg->vlan_mode) {
732 if (!strcmp(cfg->vlan_mode, "access")) {
733 s.vlan_mode = PORT_VLAN_ACCESS;
734 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
735 s.vlan_mode = PORT_VLAN_TRUNK;
736 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
737 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
738 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
739 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
741 /* This "can't happen" because ovsdb-server should prevent it. */
742 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
743 s.vlan_mode = PORT_VLAN_TRUNK;
747 s.vlan_mode = PORT_VLAN_ACCESS;
749 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
753 s.vlan_mode = PORT_VLAN_TRUNK;
756 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
759 /* Get LACP settings. */
760 s.lacp = port_configure_lacp(port, &lacp_settings);
764 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
765 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
766 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
769 s.lacp_slaves = NULL;
772 /* Get bond settings. */
773 if (s.n_slaves > 1) {
774 s.bond = &bond_settings;
775 port_configure_bond(port, &bond_settings);
778 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
779 netdev_set_miimon_interval(iface->netdev, 0);
784 ofproto_bundle_register(port->bridge->ofproto, port, &s);
792 /* Pick local port hardware address and datapath ID for 'br'. */
794 bridge_configure_datapath_id(struct bridge *br)
796 uint8_t ea[ETH_ADDR_LEN];
798 struct iface *local_iface;
799 struct iface *hw_addr_iface;
802 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
803 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
805 int error = netdev_set_etheraddr(local_iface->netdev, ea);
807 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
808 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
809 "Ethernet address: %s",
810 br->name, strerror(error));
813 memcpy(br->ea, ea, ETH_ADDR_LEN);
815 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
816 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
817 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
818 ofproto_set_datapath_id(br->ofproto, dpid);
821 dpid_string = xasprintf("%016"PRIx64, dpid);
822 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
826 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
829 bridge_get_allowed_versions(struct bridge *br)
831 if (!br->cfg->n_protocols)
834 return ofputil_versions_from_strings(br->cfg->protocols,
835 br->cfg->n_protocols);
838 /* Set NetFlow configuration on 'br'. */
840 bridge_configure_netflow(struct bridge *br)
842 struct ovsrec_netflow *cfg = br->cfg->netflow;
843 struct netflow_options opts;
846 ofproto_set_netflow(br->ofproto, NULL);
850 memset(&opts, 0, sizeof opts);
852 /* Get default NetFlow configuration from datapath.
853 * Apply overrides from 'cfg'. */
854 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
855 if (cfg->engine_type) {
856 opts.engine_type = *cfg->engine_type;
858 if (cfg->engine_id) {
859 opts.engine_id = *cfg->engine_id;
862 /* Configure active timeout interval. */
863 opts.active_timeout = cfg->active_timeout;
864 if (!opts.active_timeout) {
865 opts.active_timeout = -1;
866 } else if (opts.active_timeout < 0) {
867 VLOG_WARN("bridge %s: active timeout interval set to negative "
868 "value, using default instead (%d seconds)", br->name,
869 NF_ACTIVE_TIMEOUT_DEFAULT);
870 opts.active_timeout = -1;
873 /* Add engine ID to interface number to disambiguate bridgs? */
874 opts.add_id_to_iface = cfg->add_id_to_interface;
875 if (opts.add_id_to_iface) {
876 if (opts.engine_id > 0x7f) {
877 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
878 "another vswitch, choose an engine id less than 128",
881 if (hmap_count(&br->ports) > 508) {
882 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
883 "another port when more than 508 ports are used",
889 sset_init(&opts.collectors);
890 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
893 if (ofproto_set_netflow(br->ofproto, &opts)) {
894 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
896 sset_destroy(&opts.collectors);
899 /* Set sFlow configuration on 'br'. */
901 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
903 const struct ovsrec_sflow *cfg = br->cfg->sflow;
904 struct ovsrec_controller **controllers;
905 struct ofproto_sflow_options oso;
906 size_t n_controllers;
910 ofproto_set_sflow(br->ofproto, NULL);
914 memset(&oso, 0, sizeof oso);
916 sset_init(&oso.targets);
917 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
919 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
921 oso.sampling_rate = *cfg->sampling;
924 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
926 oso.polling_interval = *cfg->polling;
929 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
931 oso.header_len = *cfg->header;
934 oso.sub_id = (*sflow_bridge_number)++;
935 oso.agent_device = cfg->agent;
937 oso.control_ip = NULL;
938 n_controllers = bridge_get_controllers(br, &controllers);
939 for (i = 0; i < n_controllers; i++) {
940 if (controllers[i]->local_ip) {
941 oso.control_ip = controllers[i]->local_ip;
945 ofproto_set_sflow(br->ofproto, &oso);
947 sset_destroy(&oso.targets);
950 /* Set IPFIX configuration on 'br'. */
952 bridge_configure_ipfix(struct bridge *br)
954 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
955 const struct ovsrec_flow_sample_collector_set *fe_cfg;
956 struct ofproto_ipfix_bridge_exporter_options be_opts;
957 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
958 size_t n_fe_opts = 0;
960 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
961 if (fe_cfg->bridge == br->cfg) {
966 if (!be_cfg && n_fe_opts == 0) {
967 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
972 memset(&be_opts, 0, sizeof be_opts);
974 sset_init(&be_opts.targets);
975 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
977 if (be_cfg->sampling) {
978 be_opts.sampling_rate = *be_cfg->sampling;
980 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
982 if (be_cfg->obs_domain_id) {
983 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
985 if (be_cfg->obs_point_id) {
986 be_opts.obs_point_id = *be_cfg->obs_point_id;
991 struct ofproto_ipfix_flow_exporter_options *opts;
992 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
994 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
995 if (fe_cfg->bridge == br->cfg) {
996 opts->collector_set_id = fe_cfg->id;
997 sset_init(&opts->targets);
998 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
999 fe_cfg->ipfix->n_targets);
1005 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1009 sset_destroy(&be_opts.targets);
1012 if (n_fe_opts > 0) {
1013 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1015 for (i = 0; i < n_fe_opts; i++) {
1016 sset_destroy(&opts->targets);
1024 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1025 struct ofproto_port_stp_settings *port_s,
1026 int *port_num_counter, unsigned long *port_num_bitmap)
1028 const char *config_str;
1029 struct iface *iface;
1031 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1032 port_s->enable = false;
1035 port_s->enable = true;
1038 /* STP over bonds is not supported. */
1039 if (!list_is_singleton(&port->ifaces)) {
1040 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1042 port_s->enable = false;
1046 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1048 /* Internal ports shouldn't participate in spanning tree, so
1050 if (!strcmp(iface->type, "internal")) {
1051 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1052 port_s->enable = false;
1056 /* STP on mirror output ports is not supported. */
1057 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1058 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1059 port_s->enable = false;
1063 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1065 unsigned long int port_num = strtoul(config_str, NULL, 0);
1066 int port_idx = port_num - 1;
1068 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1069 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1070 port_s->enable = false;
1074 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1075 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1076 port->name, port_num);
1077 port_s->enable = false;
1080 bitmap_set1(port_num_bitmap, port_idx);
1081 port_s->port_num = port_idx;
1083 if (*port_num_counter >= STP_MAX_PORTS) {
1084 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1085 port_s->enable = false;
1089 port_s->port_num = (*port_num_counter)++;
1092 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1094 port_s->path_cost = strtoul(config_str, NULL, 10);
1096 enum netdev_features current;
1099 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1100 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1101 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1104 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1106 port_s->priority = strtoul(config_str, NULL, 0);
1108 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1112 /* Set spanning tree configuration on 'br'. */
1114 bridge_configure_stp(struct bridge *br)
1116 if (!br->cfg->stp_enable) {
1117 ofproto_set_stp(br->ofproto, NULL);
1119 struct ofproto_stp_settings br_s;
1120 const char *config_str;
1122 int port_num_counter;
1123 unsigned long *port_num_bitmap;
1125 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1127 uint8_t ea[ETH_ADDR_LEN];
1129 if (eth_addr_from_string(config_str, ea)) {
1130 br_s.system_id = eth_addr_to_uint64(ea);
1132 br_s.system_id = eth_addr_to_uint64(br->ea);
1133 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1134 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1137 br_s.system_id = eth_addr_to_uint64(br->ea);
1140 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1142 br_s.priority = strtoul(config_str, NULL, 0);
1144 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1147 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1149 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1151 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1154 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1156 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1158 br_s.max_age = STP_DEFAULT_MAX_AGE;
1161 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1163 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1165 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1168 /* Configure STP on the bridge. */
1169 if (ofproto_set_stp(br->ofproto, &br_s)) {
1170 VLOG_ERR("bridge %s: could not enable STP", br->name);
1174 /* Users must either set the port number with the "stp-port-num"
1175 * configuration on all ports or none. If manual configuration
1176 * is not done, then we allocate them sequentially. */
1177 port_num_counter = 0;
1178 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1179 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1180 struct ofproto_port_stp_settings port_s;
1181 struct iface *iface;
1183 port_configure_stp(br->ofproto, port, &port_s,
1184 &port_num_counter, port_num_bitmap);
1186 /* As bonds are not supported, just apply configuration to
1187 * all interfaces. */
1188 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1189 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1191 VLOG_ERR("port %s: could not enable STP", port->name);
1197 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1198 && port_num_counter) {
1199 VLOG_ERR("bridge %s: must manually configure all STP port "
1200 "IDs or none, disabling", br->name);
1201 ofproto_set_stp(br->ofproto, NULL);
1203 bitmap_free(port_num_bitmap);
1208 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1210 const struct port *port = port_lookup(br, name);
1211 return port && port_is_bond_fake_iface(port);
1215 port_is_bond_fake_iface(const struct port *port)
1217 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1221 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1223 struct bridge *br, *next;
1224 struct shash new_br;
1227 /* Collect new bridges' names and types. */
1228 shash_init(&new_br);
1229 for (i = 0; i < cfg->n_bridges; i++) {
1230 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1231 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1233 if (strchr(br_cfg->name, '/')) {
1234 /* Prevent remote ovsdb-server users from accessing arbitrary
1235 * directories, e.g. consider a bridge named "../../../etc/". */
1236 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1238 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1239 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1243 /* Get rid of deleted bridges or those whose types have changed.
1244 * Update 'cfg' of bridges that still exist. */
1245 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1246 br->cfg = shash_find_data(&new_br, br->name);
1247 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1248 br->cfg->datapath_type))) {
1253 /* Add new bridges. */
1254 for (i = 0; i < cfg->n_bridges; i++) {
1255 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1256 struct bridge *br = bridge_lookup(br_cfg->name);
1258 bridge_create(br_cfg);
1262 shash_destroy(&new_br);
1266 iface_set_ofp_port(struct iface *iface, int ofp_port)
1268 struct bridge *br = iface->port->bridge;
1270 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1271 iface->ofp_port = ofp_port;
1272 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1273 iface_set_ofport(iface->cfg, ofp_port);
1276 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1277 * Returns 0 if successful, otherwise a positive errno value. */
1279 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1280 struct netdev *netdev)
1284 error = netdev_set_config(netdev, &iface_cfg->options);
1286 VLOG_WARN("could not configure network device %s (%s)",
1287 iface_cfg->name, strerror(error));
1292 /* This function determines whether 'ofproto_port', which is attached to
1293 * br->ofproto's datapath, is one that we want in 'br'.
1295 * If it is, it returns true, after creating an iface (if necessary),
1296 * configuring the iface's netdev according to the iface's options, and setting
1297 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1299 * If, on the other hand, 'port' should be removed, it returns false. The
1300 * caller should later detach the port from br->ofproto. */
1302 bridge_refresh_one_ofp_port(struct bridge *br,
1303 const struct ofproto_port *ofproto_port)
1305 const char *name = ofproto_port->name;
1306 const char *type = ofproto_port->type;
1307 uint16_t ofp_port = ofproto_port->ofp_port;
1309 struct iface *iface = iface_lookup(br, name);
1311 /* Check that the name-to-number mapping is one-to-one. */
1312 if (iface->ofp_port >= 0) {
1313 VLOG_WARN("bridge %s: interface %s reported twice",
1316 } else if (iface_from_ofp_port(br, ofp_port)) {
1317 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1318 br->name, ofp_port);
1322 /* There's a configured interface named 'name'. */
1323 if (strcmp(type, iface->type)
1324 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1325 /* It's the wrong type, or it's the right type but can't be
1326 * configured as the user requested, so we must destroy it. */
1329 /* It's the right type and configured correctly. Keep it. */
1330 iface_set_ofp_port(iface, ofp_port);
1333 } else if (bridge_has_bond_fake_iface(br, name)
1334 && !strcmp(type, "internal")) {
1335 /* It's a bond fake iface. Keep it. */
1338 /* There's no configured interface named 'name', but there might be an
1339 * interface of that name queued to be created.
1341 * If there is, and it has the correct type, then try to configure it
1342 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1343 * delete it and later try to re-add it. */
1344 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1346 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1347 && iface_create(br, if_cfg, ofp_port));
1351 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1352 * consistent with the ofp_ports in "br->ofproto". */
1354 bridge_refresh_ofp_port(struct bridge *br)
1356 struct ofproto_port_dump dump;
1357 struct ofproto_port ofproto_port;
1358 struct port *port, *port_next;
1360 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1361 hmap_clear(&br->ifaces);
1362 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1363 struct iface *iface;
1365 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1366 iface->ofp_port = -1;
1370 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1371 * already exist in the datapath and promote them to full fledged "struct
1372 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1373 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1374 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1375 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1376 garbage->ofp_port = ofproto_port.ofp_port;
1377 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1381 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1382 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1383 * later they can be added to ofproto. */
1384 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1385 struct iface *iface, *iface_next;
1387 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1388 if (iface->ofp_port < 0) {
1389 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1390 iface_destroy(iface);
1394 if (list_is_empty(&port->ifaces)) {
1400 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1401 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1402 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1405 * If successful, returns 0 and stores the network device in '*netdevp'. On
1406 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1408 iface_do_create(const struct bridge *br,
1409 const struct if_cfg *if_cfg,
1410 int *ofp_portp, struct netdev **netdevp)
1412 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1413 const struct ovsrec_port *port_cfg = if_cfg->parent;
1414 struct netdev *netdev;
1417 error = netdev_open(iface_cfg->name,
1418 iface_get_type(iface_cfg, br->cfg), &netdev);
1420 VLOG_WARN("could not open network device %s (%s)",
1421 iface_cfg->name, strerror(error));
1425 error = iface_set_netdev_config(iface_cfg, netdev);
1430 if (*ofp_portp < 0) {
1431 uint16_t ofp_port = if_cfg->ofport;
1433 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1437 *ofp_portp = ofp_port;
1439 VLOG_INFO("bridge %s: added interface %s on port %d",
1440 br->name, iface_cfg->name, *ofp_portp);
1442 VLOG_DBG("bridge %s: interface %s is on port %d",
1443 br->name, iface_cfg->name, *ofp_portp);
1446 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1447 || iface_is_internal(iface_cfg, br->cfg)) {
1448 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1456 netdev_close(netdev);
1460 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1461 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1462 * automatically allocated for the iface. Takes ownership of and
1463 * deallocates 'if_cfg'.
1465 * Return true if an iface is successfully created, false otherwise. */
1467 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1469 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1470 const struct ovsrec_port *port_cfg = if_cfg->parent;
1472 struct netdev *netdev;
1473 struct iface *iface;
1478 /* Do the bits that can fail up front.
1480 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1481 * internal datastructures may not be consistent. Eventually, when port
1482 * additions and deletions are cheaper, these calls should be removed. */
1484 ovs_assert(!iface_lookup(br, iface_cfg->name));
1485 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1488 iface_set_ofport(iface_cfg, -1);
1489 iface_clear_db_record(iface_cfg);
1494 /* Get or create the port structure. */
1495 port = port_lookup(br, port_cfg->name);
1497 port = port_create(br, port_cfg);
1500 /* Create the iface structure. */
1501 iface = xzalloc(sizeof *iface);
1502 list_push_back(&port->ifaces, &iface->port_elem);
1503 hmap_insert(&br->iface_by_name, &iface->name_node,
1504 hash_string(iface_cfg->name, 0));
1506 iface->name = xstrdup(iface_cfg->name);
1507 iface->ofp_port = -1;
1508 iface->netdev = netdev;
1509 iface->type = iface_get_type(iface_cfg, br->cfg);
1510 iface->cfg = iface_cfg;
1512 iface_set_ofp_port(iface, ofp_port);
1514 /* Populate initial status in database. */
1515 iface_refresh_stats(iface);
1516 iface_refresh_status(iface);
1518 /* Add bond fake iface if necessary. */
1519 if (port_is_bond_fake_iface(port)) {
1520 struct ofproto_port ofproto_port;
1522 if (ofproto_port_query_by_name(br->ofproto, port->name,
1524 struct netdev *netdev;
1527 error = netdev_open(port->name, "internal", &netdev);
1529 uint16_t fake_ofp_port = if_cfg->ofport;
1531 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1532 netdev_close(netdev);
1534 VLOG_WARN("could not open network device %s (%s)",
1535 port->name, strerror(error));
1538 /* Already exists, nothing to do. */
1539 ofproto_port_destroy(&ofproto_port);
1544 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1550 /* Set Flow eviction threshold */
1552 bridge_configure_flow_eviction_threshold(struct bridge *br)
1554 const char *threshold_str;
1557 threshold_str = smap_get(&br->cfg->other_config,
1558 "flow-eviction-threshold");
1559 if (threshold_str) {
1560 threshold = strtoul(threshold_str, NULL, 10);
1562 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1564 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1567 /* Set forward BPDU option. */
1569 bridge_configure_forward_bpdu(struct bridge *br)
1571 ofproto_set_forward_bpdu(br->ofproto,
1572 smap_get_bool(&br->cfg->other_config,
1577 /* Set MAC learning table configuration for 'br'. */
1579 bridge_configure_mac_table(struct bridge *br)
1581 const char *idle_time_str;
1584 const char *mac_table_size_str;
1587 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1588 idle_time = (idle_time_str && atoi(idle_time_str)
1589 ? atoi(idle_time_str)
1590 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1592 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1593 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1594 ? atoi(mac_table_size_str)
1597 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1601 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1602 struct iface **hw_addr_iface)
1604 struct hmapx mirror_output_ports;
1607 bool found_addr = false;
1611 *hw_addr_iface = NULL;
1613 /* Did the user request a particular MAC? */
1614 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1615 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1616 if (eth_addr_is_multicast(ea)) {
1617 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1618 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1619 } else if (eth_addr_is_zero(ea)) {
1620 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1626 /* Mirror output ports don't participate in picking the local hardware
1627 * address. ofproto can't help us find out whether a given port is a
1628 * mirror output because we haven't configured mirrors yet, so we need to
1629 * accumulate them ourselves. */
1630 hmapx_init(&mirror_output_ports);
1631 for (i = 0; i < br->cfg->n_mirrors; i++) {
1632 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1633 if (m->output_port) {
1634 hmapx_add(&mirror_output_ports, m->output_port);
1638 /* Otherwise choose the minimum non-local MAC address among all of the
1640 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1641 uint8_t iface_ea[ETH_ADDR_LEN];
1642 struct iface *candidate;
1643 struct iface *iface;
1645 /* Mirror output ports don't participate. */
1646 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1650 /* Choose the MAC address to represent the port. */
1652 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1653 /* Find the interface with this Ethernet address (if any) so that
1654 * we can provide the correct devname to the caller. */
1655 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1656 uint8_t candidate_ea[ETH_ADDR_LEN];
1657 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1658 && eth_addr_equals(iface_ea, candidate_ea)) {
1663 /* Choose the interface whose MAC address will represent the port.
1664 * The Linux kernel bonding code always chooses the MAC address of
1665 * the first slave added to a bond, and the Fedora networking
1666 * scripts always add slaves to a bond in alphabetical order, so
1667 * for compatibility we choose the interface with the name that is
1668 * first in alphabetical order. */
1669 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1670 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1675 /* The local port doesn't count (since we're trying to choose its
1676 * MAC address anyway). */
1677 if (iface->ofp_port == OFPP_LOCAL) {
1682 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1688 /* Compare against our current choice. */
1689 if (!eth_addr_is_multicast(iface_ea) &&
1690 !eth_addr_is_local(iface_ea) &&
1691 !eth_addr_is_reserved(iface_ea) &&
1692 !eth_addr_is_zero(iface_ea) &&
1693 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1695 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1696 *hw_addr_iface = iface;
1702 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1703 *hw_addr_iface = NULL;
1706 hmapx_destroy(&mirror_output_ports);
1709 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1710 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1711 * an interface on 'br', then that interface must be passed in as
1712 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1713 * 'hw_addr_iface' must be passed in as a null pointer. */
1715 bridge_pick_datapath_id(struct bridge *br,
1716 const uint8_t bridge_ea[ETH_ADDR_LEN],
1717 struct iface *hw_addr_iface)
1720 * The procedure for choosing a bridge MAC address will, in the most
1721 * ordinary case, also choose a unique MAC that we can use as a datapath
1722 * ID. In some special cases, though, multiple bridges will end up with
1723 * the same MAC address. This is OK for the bridges, but it will confuse
1724 * the OpenFlow controller, because each datapath needs a unique datapath
1727 * Datapath IDs must be unique. It is also very desirable that they be
1728 * stable from one run to the next, so that policy set on a datapath
1731 const char *datapath_id;
1734 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1735 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1739 if (!hw_addr_iface) {
1741 * A purely internal bridge, that is, one that has no non-virtual
1742 * network devices on it at all, is difficult because it has no
1743 * natural unique identifier at all.
1745 * When the host is a XenServer, we handle this case by hashing the
1746 * host's UUID with the name of the bridge. Names of bridges are
1747 * persistent across XenServer reboots, although they can be reused if
1748 * an internal network is destroyed and then a new one is later
1749 * created, so this is fairly effective.
1751 * When the host is not a XenServer, we punt by using a random MAC
1752 * address on each run.
1754 const char *host_uuid = xenserver_get_host_uuid();
1756 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1757 dpid = dpid_from_hash(combined, strlen(combined));
1763 return eth_addr_to_uint64(bridge_ea);
1767 dpid_from_hash(const void *data, size_t n)
1769 uint8_t hash[SHA1_DIGEST_SIZE];
1771 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1772 sha1_bytes(data, n, hash);
1773 eth_addr_mark_random(hash);
1774 return eth_addr_to_uint64(hash);
1778 iface_refresh_status(struct iface *iface)
1782 enum netdev_features current;
1786 uint8_t mac[ETH_ADDR_LEN];
1789 if (iface_is_synthetic(iface)) {
1795 if (!netdev_get_status(iface->netdev, &smap)) {
1796 ovsrec_interface_set_status(iface->cfg, &smap);
1798 ovsrec_interface_set_status(iface->cfg, NULL);
1801 smap_destroy(&smap);
1803 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1804 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1806 ovsrec_interface_set_duplex(iface->cfg,
1807 netdev_features_is_full_duplex(current)
1809 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1811 ovsrec_interface_set_duplex(iface->cfg, NULL);
1812 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1815 error = netdev_get_mtu(iface->netdev, &mtu);
1818 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1820 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1823 error = netdev_get_etheraddr(iface->netdev, mac);
1825 char mac_string[32];
1827 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1828 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1830 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1834 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1837 iface_refresh_cfm_stats(struct iface *iface)
1839 const struct ovsrec_interface *cfg = iface->cfg;
1840 struct ofproto_cfm_status status;
1842 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1843 iface->ofp_port, &status)) {
1844 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1845 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1846 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1847 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1848 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1850 const char *reasons[CFM_FAULT_N_REASONS];
1851 int64_t cfm_health = status.health;
1852 bool faulted = status.faults != 0;
1855 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1858 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1859 int reason = 1 << i;
1860 if (status.faults & reason) {
1861 reasons[j++] = cfm_fault_reason_to_str(reason);
1864 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1866 if (status.remote_opstate >= 0) {
1867 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1868 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1870 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1873 ovsrec_interface_set_cfm_remote_mpids(cfg,
1874 (const int64_t *)status.rmps,
1876 if (cfm_health >= 0) {
1877 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1879 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1885 iface_refresh_stats(struct iface *iface)
1887 #define IFACE_STATS \
1888 IFACE_STAT(rx_packets, "rx_packets") \
1889 IFACE_STAT(tx_packets, "tx_packets") \
1890 IFACE_STAT(rx_bytes, "rx_bytes") \
1891 IFACE_STAT(tx_bytes, "tx_bytes") \
1892 IFACE_STAT(rx_dropped, "rx_dropped") \
1893 IFACE_STAT(tx_dropped, "tx_dropped") \
1894 IFACE_STAT(rx_errors, "rx_errors") \
1895 IFACE_STAT(tx_errors, "tx_errors") \
1896 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1897 IFACE_STAT(rx_over_errors, "rx_over_err") \
1898 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1899 IFACE_STAT(collisions, "collisions")
1901 #define IFACE_STAT(MEMBER, NAME) NAME,
1902 static char *keys[] = { IFACE_STATS };
1904 int64_t values[ARRAY_SIZE(keys)];
1907 struct netdev_stats stats;
1909 if (iface_is_synthetic(iface)) {
1913 /* Intentionally ignore return value, since errors will set 'stats' to
1914 * all-1s, and we will deal with that correctly below. */
1915 netdev_get_stats(iface->netdev, &stats);
1917 /* Copy statistics into values[] array. */
1919 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1922 ovs_assert(i == ARRAY_SIZE(keys));
1924 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1930 br_refresh_stp_status(struct bridge *br)
1932 struct smap smap = SMAP_INITIALIZER(&smap);
1933 struct ofproto *ofproto = br->ofproto;
1934 struct ofproto_stp_status status;
1936 if (ofproto_get_stp_status(ofproto, &status)) {
1940 if (!status.enabled) {
1941 ovsrec_bridge_set_status(br->cfg, NULL);
1945 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1946 STP_ID_ARGS(status.bridge_id));
1947 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1948 STP_ID_ARGS(status.designated_root));
1949 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1951 ovsrec_bridge_set_status(br->cfg, &smap);
1952 smap_destroy(&smap);
1956 port_refresh_stp_status(struct port *port)
1958 struct ofproto *ofproto = port->bridge->ofproto;
1959 struct iface *iface;
1960 struct ofproto_port_stp_status status;
1962 int64_t int_values[3];
1965 if (port_is_synthetic(port)) {
1969 /* STP doesn't currently support bonds. */
1970 if (!list_is_singleton(&port->ifaces)) {
1971 ovsrec_port_set_status(port->cfg, NULL);
1975 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1977 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1981 if (!status.enabled) {
1982 ovsrec_port_set_status(port->cfg, NULL);
1983 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1987 /* Set Status column. */
1989 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1990 smap_add(&smap, "stp_state", stp_state_name(status.state));
1991 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1992 smap_add(&smap, "stp_role", stp_role_name(status.role));
1993 ovsrec_port_set_status(port->cfg, &smap);
1994 smap_destroy(&smap);
1996 /* Set Statistics column. */
1997 keys[0] = "stp_tx_count";
1998 int_values[0] = status.tx_count;
1999 keys[1] = "stp_rx_count";
2000 int_values[1] = status.rx_count;
2001 keys[2] = "stp_error_count";
2002 int_values[2] = status.error_count;
2004 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2005 ARRAY_SIZE(int_values));
2009 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2011 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2015 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2017 bool enable = enable_system_stats(cfg);
2019 system_stats_enable(enable);
2021 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2026 run_system_stats(void)
2028 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2031 stats = system_stats_run();
2033 struct ovsdb_idl_txn *txn;
2034 struct ovsdb_datum datum;
2036 txn = ovsdb_idl_txn_create(idl);
2037 ovsdb_datum_from_smap(&datum, stats);
2038 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2040 ovsdb_idl_txn_commit(txn);
2041 ovsdb_idl_txn_destroy(txn);
2047 static inline const char *
2048 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2051 case OFPCR12_ROLE_EQUAL:
2053 case OFPCR12_ROLE_MASTER:
2055 case OFPCR12_ROLE_SLAVE:
2057 case OFPCR12_ROLE_NOCHANGE:
2059 return "*** INVALID ROLE ***";
2064 refresh_controller_status(void)
2068 const struct ovsrec_controller *cfg;
2072 /* Accumulate status for controllers on all bridges. */
2073 HMAP_FOR_EACH (br, node, &all_bridges) {
2074 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2077 /* Update each controller in the database with current status. */
2078 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2079 struct ofproto_controller_info *cinfo =
2080 shash_find_data(&info, cfg->target);
2083 struct smap smap = SMAP_INITIALIZER(&smap);
2084 const char **values = cinfo->pairs.values;
2085 const char **keys = cinfo->pairs.keys;
2088 for (i = 0; i < cinfo->pairs.n; i++) {
2089 smap_add(&smap, keys[i], values[i]);
2092 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2093 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2095 ovsrec_controller_set_status(cfg, &smap);
2096 smap_destroy(&smap);
2098 ovsrec_controller_set_is_connected(cfg, false);
2099 ovsrec_controller_set_role(cfg, NULL);
2100 ovsrec_controller_set_status(cfg, NULL);
2104 ofproto_free_ofproto_controller_info(&info);
2109 * Some information in the database must be kept as up-to-date as possible to
2110 * allow controllers to respond rapidly to network outages. We call these
2111 * statistics "instant" stats.
2113 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2114 * assuming that they've changed. The only means we have to determine whether
2115 * they have changed are:
2117 * - Try to commit changes to the database. If nothing changed, then
2118 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2121 * - instant_stats_run() is called late in the run loop, after anything that
2122 * might change any of the instant stats.
2124 * We use these two facts together to avoid waking the process up every
2125 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2128 /* Minimum interval between writing updates to the instant stats to the
2130 #define INSTANT_INTERVAL_MSEC 100
2132 /* Current instant stats database transaction, NULL if there is no ongoing
2134 static struct ovsdb_idl_txn *instant_txn;
2136 /* Next time (in msec on monotonic clock) at which we will update the instant
2138 static long long int instant_next_txn = LLONG_MIN;
2140 /* True if the run loop has run since we last saw that the instant stats were
2141 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2142 * to refresh the instant stats. */
2143 static bool instant_stats_could_have_changed;
2146 instant_stats_run(void)
2148 enum ovsdb_idl_txn_status status;
2150 instant_stats_could_have_changed = true;
2155 if (time_msec() < instant_next_txn) {
2158 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2160 instant_txn = ovsdb_idl_txn_create(idl);
2161 HMAP_FOR_EACH (br, node, &all_bridges) {
2162 struct iface *iface;
2165 br_refresh_stp_status(br);
2167 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2168 port_refresh_stp_status(port);
2171 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2172 enum netdev_flags flags;
2173 const char *link_state;
2174 int64_t link_resets;
2177 if (iface_is_synthetic(iface)) {
2181 current = ofproto_port_is_lacp_current(br->ofproto,
2185 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2187 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2190 error = netdev_get_flags(iface->netdev, &flags);
2192 const char *state = flags & NETDEV_UP ? "up" : "down";
2193 ovsrec_interface_set_admin_state(iface->cfg, state);
2195 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2198 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2199 ovsrec_interface_set_link_state(iface->cfg, link_state);
2201 link_resets = netdev_get_carrier_resets(iface->netdev);
2202 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2204 iface_refresh_cfm_stats(iface);
2209 status = ovsdb_idl_txn_commit(instant_txn);
2210 if (status != TXN_INCOMPLETE) {
2211 ovsdb_idl_txn_destroy(instant_txn);
2214 if (status == TXN_UNCHANGED) {
2215 instant_stats_could_have_changed = false;
2220 instant_stats_wait(void)
2223 ovsdb_idl_txn_wait(instant_txn);
2224 } else if (instant_stats_could_have_changed) {
2225 poll_timer_wait_until(instant_next_txn);
2229 /* Performs periodic activity required by bridges that needs to be done with
2230 * the least possible latency.
2232 * It makes sense to call this function a couple of times per poll loop, to
2233 * provide a significant performance boost on some benchmarks with ofprotos
2234 * that use the ofproto-dpif implementation. */
2236 bridge_run_fast(void)
2243 ofproto_enumerate_types(&types);
2244 SSET_FOR_EACH (type, &types) {
2245 ofproto_type_run_fast(type);
2247 sset_destroy(&types);
2249 HMAP_FOR_EACH (br, node, &all_bridges) {
2250 ofproto_run_fast(br->ofproto);
2257 static struct ovsrec_open_vswitch null_cfg;
2258 const struct ovsrec_open_vswitch *cfg;
2259 struct ovsdb_idl_txn *reconf_txn = NULL;
2263 bool vlan_splinters_changed;
2266 ovsrec_open_vswitch_init(&null_cfg);
2268 /* (Re)configure if necessary. */
2269 if (!reconfiguring) {
2272 if (ovsdb_idl_is_lock_contended(idl)) {
2273 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2274 struct bridge *br, *next_br;
2276 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2277 "disabling this process (pid %ld) until it goes away",
2278 (long int) getpid());
2280 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2283 /* Since we will not be running system_stats_run() in this process
2284 * with the current situation of multiple ovs-vswitchd daemons,
2285 * disable system stats collection. */
2286 system_stats_enable(false);
2288 } else if (!ovsdb_idl_has_lock(idl)) {
2292 cfg = ovsrec_open_vswitch_first(idl);
2294 /* Initialize the ofproto library. This only needs to run once, but
2295 * it must be done after the configuration is set. If the
2296 * initialization has already occurred, bridge_init_ofproto()
2297 * returns immediately. */
2298 bridge_init_ofproto(cfg);
2300 /* Let each datapath type do the work that it needs to do. */
2302 ofproto_enumerate_types(&types);
2303 SSET_FOR_EACH (type, &types) {
2304 ofproto_type_run(type);
2306 sset_destroy(&types);
2308 /* Let each bridge do the work that it needs to do. */
2309 HMAP_FOR_EACH (br, node, &all_bridges) {
2310 ofproto_run(br->ofproto);
2313 /* Re-configure SSL. We do this on every trip through the main loop,
2314 * instead of just when the database changes, because the contents of the
2315 * key and certificate files can change without the database changing.
2317 * We do this before bridge_reconfigure() because that function might
2318 * initiate SSL connections and thus requires SSL to be configured. */
2319 if (cfg && cfg->ssl) {
2320 const struct ovsrec_ssl *ssl = cfg->ssl;
2322 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2323 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2326 if (!reconfiguring) {
2327 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2328 * usage has changed. */
2329 vlan_splinters_changed = false;
2330 if (vlan_splinters_enabled_anywhere) {
2331 HMAP_FOR_EACH (br, node, &all_bridges) {
2332 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2333 vlan_splinters_changed = true;
2339 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2340 idl_seqno = ovsdb_idl_get_seqno(idl);
2342 reconf_txn = ovsdb_idl_txn_create(idl);
2343 bridge_reconfigure(cfg);
2345 /* We still need to reconfigure to avoid dangling pointers to
2346 * now-destroyed ovsrec structures inside bridge data. */
2347 bridge_reconfigure(&null_cfg);
2352 if (reconfiguring) {
2354 reconf_txn = ovsdb_idl_txn_create(idl);
2357 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2358 reconfiguring = false;
2361 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2364 /* If we are completing our initial configuration for this run
2365 * of ovs-vswitchd, then keep the transaction around to monitor
2366 * it for completion. */
2367 if (!initial_config_done) {
2368 initial_config_done = true;
2369 daemonize_txn = reconf_txn;
2376 ovsdb_idl_txn_commit(reconf_txn);
2377 ovsdb_idl_txn_destroy(reconf_txn);
2381 if (daemonize_txn) {
2382 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2383 if (status != TXN_INCOMPLETE) {
2384 ovsdb_idl_txn_destroy(daemonize_txn);
2385 daemonize_txn = NULL;
2387 /* ovs-vswitchd has completed initialization, so allow the
2388 * process that forked us to exit successfully. */
2389 daemonize_complete();
2391 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2395 /* Refresh interface and mirror stats if necessary. */
2396 if (time_msec() >= iface_stats_timer) {
2398 struct ovsdb_idl_txn *txn;
2400 txn = ovsdb_idl_txn_create(idl);
2401 HMAP_FOR_EACH (br, node, &all_bridges) {
2405 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2406 struct iface *iface;
2408 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2409 iface_refresh_stats(iface);
2410 iface_refresh_status(iface);
2414 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2415 mirror_refresh_stats(m);
2419 refresh_controller_status();
2420 ovsdb_idl_txn_commit(txn);
2421 ovsdb_idl_txn_destroy(txn); /* XXX */
2424 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2428 instant_stats_run();
2437 ovsdb_idl_wait(idl);
2438 if (daemonize_txn) {
2439 ovsdb_idl_txn_wait(daemonize_txn);
2442 if (reconfiguring) {
2443 poll_immediate_wake();
2447 ofproto_enumerate_types(&types);
2448 SSET_FOR_EACH (type, &types) {
2449 ofproto_type_wait(type);
2451 sset_destroy(&types);
2453 if (!hmap_is_empty(&all_bridges)) {
2456 HMAP_FOR_EACH (br, node, &all_bridges) {
2457 ofproto_wait(br->ofproto);
2459 poll_timer_wait_until(iface_stats_timer);
2462 system_stats_wait();
2463 instant_stats_wait();
2466 /* Adds some memory usage statistics for bridges into 'usage', for use with
2467 * memory_report(). */
2469 bridge_get_memory_usage(struct simap *usage)
2473 HMAP_FOR_EACH (br, node, &all_bridges) {
2474 ofproto_get_memory_usage(br->ofproto, usage);
2478 /* QoS unixctl user interface functions. */
2480 struct qos_unixctl_show_cbdata {
2482 struct iface *iface;
2486 qos_unixctl_show_cb(unsigned int queue_id,
2487 const struct smap *details,
2490 struct qos_unixctl_show_cbdata *data = aux;
2491 struct ds *ds = data->ds;
2492 struct iface *iface = data->iface;
2493 struct netdev_queue_stats stats;
2494 struct smap_node *node;
2497 ds_put_cstr(ds, "\n");
2499 ds_put_format(ds, "Queue %u:\n", queue_id);
2501 ds_put_cstr(ds, "Default:\n");
2504 SMAP_FOR_EACH (node, details) {
2505 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2508 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2510 if (stats.tx_packets != UINT64_MAX) {
2511 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2514 if (stats.tx_bytes != UINT64_MAX) {
2515 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2518 if (stats.tx_errors != UINT64_MAX) {
2519 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2522 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2523 queue_id, strerror(error));
2528 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2529 const char *argv[], void *aux OVS_UNUSED)
2531 struct ds ds = DS_EMPTY_INITIALIZER;
2532 struct smap smap = SMAP_INITIALIZER(&smap);
2533 struct iface *iface;
2535 struct smap_node *node;
2536 struct qos_unixctl_show_cbdata data;
2539 iface = iface_find(argv[1]);
2541 unixctl_command_reply_error(conn, "no such interface");
2545 netdev_get_qos(iface->netdev, &type, &smap);
2547 if (*type != '\0') {
2548 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2550 SMAP_FOR_EACH (node, &smap) {
2551 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2556 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2559 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2561 unixctl_command_reply(conn, ds_cstr(&ds));
2563 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2564 unixctl_command_reply_error(conn, ds_cstr(&ds));
2567 smap_destroy(&smap);
2571 /* Bridge reconfiguration functions. */
2573 bridge_create(const struct ovsrec_bridge *br_cfg)
2577 ovs_assert(!bridge_lookup(br_cfg->name));
2578 br = xzalloc(sizeof *br);
2580 br->name = xstrdup(br_cfg->name);
2581 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2584 /* Derive the default Ethernet address from the bridge's UUID. This should
2585 * be unique and it will be stable between ovs-vswitchd runs. */
2586 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2587 eth_addr_mark_random(br->default_ea);
2589 hmap_init(&br->ports);
2590 hmap_init(&br->ifaces);
2591 hmap_init(&br->iface_by_name);
2592 hmap_init(&br->mirrors);
2594 hmap_init(&br->if_cfg_todo);
2595 list_init(&br->ofpp_garbage);
2597 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2601 bridge_destroy(struct bridge *br)
2604 struct mirror *mirror, *next_mirror;
2605 struct port *port, *next_port;
2606 struct if_cfg *if_cfg, *next_if_cfg;
2607 struct ofpp_garbage *garbage, *next_garbage;
2609 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2612 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2613 mirror_destroy(mirror);
2615 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2616 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2619 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2620 &br->ofpp_garbage) {
2621 list_remove(&garbage->list_node);
2625 hmap_remove(&all_bridges, &br->node);
2626 ofproto_destroy(br->ofproto);
2627 hmap_destroy(&br->ifaces);
2628 hmap_destroy(&br->ports);
2629 hmap_destroy(&br->iface_by_name);
2630 hmap_destroy(&br->mirrors);
2631 hmap_destroy(&br->if_cfg_todo);
2638 static struct bridge *
2639 bridge_lookup(const char *name)
2643 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2644 if (!strcmp(br->name, name)) {
2651 /* Handle requests for a listing of all flows known by the OpenFlow
2652 * stack, including those normally hidden. */
2654 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2655 const char *argv[], void *aux OVS_UNUSED)
2660 br = bridge_lookup(argv[1]);
2662 unixctl_command_reply_error(conn, "Unknown bridge");
2667 ofproto_get_all_flows(br->ofproto, &results);
2669 unixctl_command_reply(conn, ds_cstr(&results));
2670 ds_destroy(&results);
2673 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2674 * connections and reconnect. If BRIDGE is not specified, then all bridges
2675 * drop their controller connections and reconnect. */
2677 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2678 const char *argv[], void *aux OVS_UNUSED)
2682 br = bridge_lookup(argv[1]);
2684 unixctl_command_reply_error(conn, "Unknown bridge");
2687 ofproto_reconnect_controllers(br->ofproto);
2689 HMAP_FOR_EACH (br, node, &all_bridges) {
2690 ofproto_reconnect_controllers(br->ofproto);
2693 unixctl_command_reply(conn, NULL);
2697 bridge_get_controllers(const struct bridge *br,
2698 struct ovsrec_controller ***controllersp)
2700 struct ovsrec_controller **controllers;
2701 size_t n_controllers;
2703 controllers = br->cfg->controller;
2704 n_controllers = br->cfg->n_controller;
2706 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2712 *controllersp = controllers;
2714 return n_controllers;
2718 bridge_queue_if_cfg(struct bridge *br,
2719 const struct ovsrec_interface *cfg,
2720 const struct ovsrec_port *parent)
2722 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2725 if_cfg->parent = parent;
2726 if_cfg->ofport = iface_pick_ofport(cfg);
2727 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2728 hash_string(if_cfg->cfg->name, 0));
2731 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2732 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2733 * 'br' needs to complete its configuration. */
2735 bridge_add_del_ports(struct bridge *br,
2736 const unsigned long int *splinter_vlans)
2738 struct shash_node *port_node;
2739 struct port *port, *next;
2740 struct shash new_ports;
2743 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2745 /* Collect new ports. */
2746 shash_init(&new_ports);
2747 for (i = 0; i < br->cfg->n_ports; i++) {
2748 const char *name = br->cfg->ports[i]->name;
2749 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2750 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2754 if (bridge_get_controllers(br, NULL)
2755 && !shash_find(&new_ports, br->name)) {
2756 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2757 br->name, br->name);
2759 ovsrec_interface_init(&br->synth_local_iface);
2760 ovsrec_port_init(&br->synth_local_port);
2762 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2763 br->synth_local_port.n_interfaces = 1;
2764 br->synth_local_port.name = br->name;
2766 br->synth_local_iface.name = br->name;
2767 br->synth_local_iface.type = "internal";
2769 br->synth_local_ifacep = &br->synth_local_iface;
2771 shash_add(&new_ports, br->name, &br->synth_local_port);
2774 if (splinter_vlans) {
2775 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2778 /* Get rid of deleted ports.
2779 * Get rid of deleted interfaces on ports that still exist. */
2780 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2781 port->cfg = shash_find_data(&new_ports, port->name);
2785 port_del_ifaces(port);
2789 /* Update iface->cfg and iface->type in interfaces that still exist.
2790 * Add new interfaces to creation queue. */
2791 SHASH_FOR_EACH (port_node, &new_ports) {
2792 const struct ovsrec_port *port = port_node->data;
2795 for (i = 0; i < port->n_interfaces; i++) {
2796 const struct ovsrec_interface *cfg = port->interfaces[i];
2797 struct iface *iface = iface_lookup(br, cfg->name);
2798 const char *type = iface_get_type(cfg, br->cfg);
2803 } else if (!strcmp(type, "null")) {
2804 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2805 " may be removed in February 2013. Please email"
2806 " dev@openvswitch.org with concerns.",
2809 bridge_queue_if_cfg(br, cfg, port);
2814 shash_destroy(&new_ports);
2817 /* Initializes 'oc' appropriately as a management service controller for
2820 * The caller must free oc->target when it is no longer needed. */
2822 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2823 struct ofproto_controller *oc)
2825 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2826 oc->max_backoff = 0;
2827 oc->probe_interval = 60;
2828 oc->band = OFPROTO_OUT_OF_BAND;
2830 oc->burst_limit = 0;
2831 oc->enable_async_msgs = true;
2834 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2836 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2837 struct ofproto_controller *oc)
2841 oc->target = c->target;
2842 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2843 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2844 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2845 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2846 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2847 oc->burst_limit = (c->controller_burst_limit
2848 ? *c->controller_burst_limit : 0);
2849 oc->enable_async_msgs = (!c->enable_async_messages
2850 || *c->enable_async_messages);
2851 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2852 if (dscp < 0 || dscp > 63) {
2853 dscp = DSCP_DEFAULT;
2858 /* Configures the IP stack for 'br''s local interface properly according to the
2859 * configuration in 'c'. */
2861 bridge_configure_local_iface_netdev(struct bridge *br,
2862 struct ovsrec_controller *c)
2864 struct netdev *netdev;
2865 struct in_addr mask, gateway;
2867 struct iface *local_iface;
2870 /* If there's no local interface or no IP address, give up. */
2871 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2872 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2876 /* Bring up the local interface. */
2877 netdev = local_iface->netdev;
2878 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2880 /* Configure the IP address and netmask. */
2881 if (!c->local_netmask
2882 || !inet_aton(c->local_netmask, &mask)
2884 mask.s_addr = guess_netmask(ip.s_addr);
2886 if (!netdev_set_in4(netdev, ip, mask)) {
2887 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2888 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2891 /* Configure the default gateway. */
2892 if (c->local_gateway
2893 && inet_aton(c->local_gateway, &gateway)
2894 && gateway.s_addr) {
2895 if (!netdev_add_router(netdev, gateway)) {
2896 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2897 br->name, IP_ARGS(gateway.s_addr));
2902 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2903 * in either string are treated as equal to any number of slashes in the other,
2904 * e.g. "x///y" is equal to "x/y".
2906 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2907 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2908 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2909 * 'b' against a prefix of 'a'.
2912 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2914 const char *b_start = b;
2916 if (b - b_start >= b_stoplen) {
2918 } else if (*a != *b) {
2920 } else if (*a == '/') {
2921 a += strspn(a, "/");
2922 b += strspn(b, "/");
2923 } else if (*a == '\0') {
2933 bridge_configure_remotes(struct bridge *br,
2934 const struct sockaddr_in *managers, size_t n_managers)
2936 bool disable_in_band;
2938 struct ovsrec_controller **controllers;
2939 size_t n_controllers;
2941 enum ofproto_fail_mode fail_mode;
2943 struct ofproto_controller *ocs;
2947 /* Check if we should disable in-band control on this bridge. */
2948 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2951 /* Set OpenFlow queue ID for in-band control. */
2952 ofproto_set_in_band_queue(br->ofproto,
2953 smap_get_int(&br->cfg->other_config,
2954 "in-band-queue", -1));
2956 if (disable_in_band) {
2957 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2959 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2962 n_controllers = bridge_get_controllers(br, &controllers);
2964 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2967 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2968 for (i = 0; i < n_controllers; i++) {
2969 struct ovsrec_controller *c = controllers[i];
2971 if (!strncmp(c->target, "punix:", 6)
2972 || !strncmp(c->target, "unix:", 5)) {
2973 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2976 if (!strncmp(c->target, "unix:", 5)) {
2977 /* Connect to a listening socket */
2978 whitelist = xasprintf("unix:%s/", ovs_rundir());
2979 if (strchr(c->target, '/') &&
2980 !equal_pathnames(c->target, whitelist,
2981 strlen(whitelist))) {
2982 /* Absolute path specified, but not in ovs_rundir */
2983 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2984 "controller \"%s\" due to possibility for "
2985 "remote exploit. Instead, specify socket "
2986 "in whitelisted \"%s\" or connect to "
2987 "\"unix:%s/%s.mgmt\" (which is always "
2988 "available without special configuration).",
2989 br->name, c->target, whitelist,
2990 ovs_rundir(), br->name);
2995 whitelist = xasprintf("punix:%s/%s.controller",
2996 ovs_rundir(), br->name);
2997 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2998 /* Prevent remote ovsdb-server users from accessing
2999 * arbitrary Unix domain sockets and overwriting arbitrary
3001 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3002 "controller \"%s\" due to possibility of "
3003 "overwriting local files. Instead, specify "
3004 "whitelisted \"%s\" or connect to "
3005 "\"unix:%s/%s.mgmt\" (which is always "
3006 "available without special configuration).",
3007 br->name, c->target, whitelist,
3008 ovs_rundir(), br->name);
3017 bridge_configure_local_iface_netdev(br, c);
3018 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3019 if (disable_in_band) {
3020 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3025 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3026 bridge_get_allowed_versions(br));
3027 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3030 /* Set the fail-mode. */
3031 fail_mode = !br->cfg->fail_mode
3032 || !strcmp(br->cfg->fail_mode, "standalone")
3033 ? OFPROTO_FAIL_STANDALONE
3034 : OFPROTO_FAIL_SECURE;
3035 ofproto_set_fail_mode(br->ofproto, fail_mode);
3037 /* Configure OpenFlow controller connection snooping. */
3038 if (!ofproto_has_snoops(br->ofproto)) {
3042 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3043 ovs_rundir(), br->name));
3044 ofproto_set_snoops(br->ofproto, &snoops);
3045 sset_destroy(&snoops);
3050 bridge_configure_tables(struct bridge *br)
3052 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3056 n_tables = ofproto_get_n_tables(br->ofproto);
3058 for (i = 0; i < n_tables; i++) {
3059 struct ofproto_table_settings s;
3062 s.max_flows = UINT_MAX;
3066 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3067 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3070 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3071 s.max_flows = *cfg->flow_limit;
3073 if (cfg->overflow_policy
3074 && !strcmp(cfg->overflow_policy, "evict")) {
3077 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3078 for (k = 0; k < cfg->n_groups; k++) {
3079 const char *string = cfg->groups[k];
3082 msg = mf_parse_subfield__(&s.groups[k], &string);
3084 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3085 "'groups' (%s)", br->name, i, msg);
3087 } else if (*string) {
3088 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3089 "element '%s' contains trailing garbage",
3090 br->name, i, cfg->groups[k]);
3098 ofproto_configure_table(br->ofproto, i, &s);
3102 for (; j < br->cfg->n_flow_tables; j++) {
3103 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3104 "%"PRId64" not supported by this datapath", br->name,
3105 br->cfg->key_flow_tables[j]);
3110 bridge_configure_dp_desc(struct bridge *br)
3112 ofproto_set_dp_desc(br->ofproto,
3113 smap_get(&br->cfg->other_config, "dp-desc"));
3116 /* Port functions. */
3118 static struct port *
3119 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3123 port = xzalloc(sizeof *port);
3125 port->name = xstrdup(cfg->name);
3127 list_init(&port->ifaces);
3129 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3133 /* Deletes interfaces from 'port' that are no longer configured for it. */
3135 port_del_ifaces(struct port *port)
3137 struct iface *iface, *next;
3138 struct sset new_ifaces;
3141 /* Collect list of new interfaces. */
3142 sset_init(&new_ifaces);
3143 for (i = 0; i < port->cfg->n_interfaces; i++) {
3144 const char *name = port->cfg->interfaces[i]->name;
3145 const char *type = port->cfg->interfaces[i]->type;
3146 if (strcmp(type, "null")) {
3147 sset_add(&new_ifaces, name);
3151 /* Get rid of deleted interfaces. */
3152 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3153 if (!sset_contains(&new_ifaces, iface->name)) {
3154 iface_destroy(iface);
3158 sset_destroy(&new_ifaces);
3162 port_destroy(struct port *port)
3165 struct bridge *br = port->bridge;
3166 struct iface *iface, *next;
3169 ofproto_bundle_unregister(br->ofproto, port);
3172 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3173 iface_destroy(iface);
3176 hmap_remove(&br->ports, &port->hmap_node);
3182 static struct port *
3183 port_lookup(const struct bridge *br, const char *name)
3187 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3189 if (!strcmp(port->name, name)) {
3197 enable_lacp(struct port *port, bool *activep)
3199 if (!port->cfg->lacp) {
3200 /* XXX when LACP implementation has been sufficiently tested, enable by
3201 * default and make active on bonded ports. */
3203 } else if (!strcmp(port->cfg->lacp, "off")) {
3205 } else if (!strcmp(port->cfg->lacp, "active")) {
3208 } else if (!strcmp(port->cfg->lacp, "passive")) {
3212 VLOG_WARN("port %s: unknown LACP mode %s",
3213 port->name, port->cfg->lacp);
3218 static struct lacp_settings *
3219 port_configure_lacp(struct port *port, struct lacp_settings *s)
3221 const char *lacp_time, *system_id;
3224 if (!enable_lacp(port, &s->active)) {
3228 s->name = port->name;
3230 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3232 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3233 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3234 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3235 " address.", port->name, system_id);
3239 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3242 if (eth_addr_is_zero(s->id)) {
3243 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3247 /* Prefer bondable links if unspecified. */
3248 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3250 s->priority = (priority > 0 && priority <= UINT16_MAX
3252 : UINT16_MAX - !list_is_short(&port->ifaces));
3254 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3255 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3260 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3262 int priority, portid, key;
3264 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3265 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3267 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3269 if (portid <= 0 || portid > UINT16_MAX) {
3270 portid = iface->ofp_port;
3273 if (priority <= 0 || priority > UINT16_MAX) {
3274 priority = UINT16_MAX;
3277 if (key < 0 || key > UINT16_MAX) {
3281 s->name = iface->name;
3283 s->priority = priority;
3288 port_configure_bond(struct port *port, struct bond_settings *s)
3290 const char *detect_s;
3291 struct iface *iface;
3292 int miimon_interval;
3294 s->name = port->name;
3296 if (port->cfg->bond_mode) {
3297 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3298 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3299 port->name, port->cfg->bond_mode,
3300 bond_mode_to_string(s->balance));
3303 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3305 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3306 * active-backup. At some point we should remove this warning. */
3307 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3308 " in previous versions, the default bond_mode was"
3309 " balance-slb", port->name,
3310 bond_mode_to_string(s->balance));
3312 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3313 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3314 "please use another bond type or disable flood_vlans",
3318 miimon_interval = smap_get_int(&port->cfg->other_config,
3319 "bond-miimon-interval", 0);
3320 if (miimon_interval <= 0) {
3321 miimon_interval = 200;
3324 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3325 if (!detect_s || !strcmp(detect_s, "carrier")) {
3326 miimon_interval = 0;
3327 } else if (strcmp(detect_s, "miimon")) {
3328 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3329 "defaulting to carrier", port->name, detect_s);
3330 miimon_interval = 0;
3333 s->up_delay = MAX(0, port->cfg->bond_updelay);
3334 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3335 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3336 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3337 "bond-rebalance-interval", 10000);
3338 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3339 s->rebalance_interval = 1000;
3342 s->fake_iface = port->cfg->bond_fake_iface;
3344 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3345 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3349 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3350 * instead of obtaining it from the database. */
3352 port_is_synthetic(const struct port *port)
3354 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3357 /* Interface functions. */
3360 iface_is_internal(const struct ovsrec_interface *iface,
3361 const struct ovsrec_bridge *br)
3363 /* The local port and "internal" ports are always "internal". */
3364 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3367 /* Returns the correct network device type for interface 'iface' in bridge
3370 iface_get_type(const struct ovsrec_interface *iface,
3371 const struct ovsrec_bridge *br)
3375 /* The local port always has type "internal". Other ports take
3376 * their type from the database and default to "system" if none is
3378 if (iface_is_internal(iface, br)) {
3381 type = iface->type[0] ? iface->type : "system";
3384 return ofproto_port_open_type(br->datapath_type, type);
3388 iface_destroy(struct iface *iface)
3391 struct port *port = iface->port;
3392 struct bridge *br = port->bridge;
3394 if (br->ofproto && iface->ofp_port >= 0) {
3395 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3398 if (iface->ofp_port >= 0) {
3399 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3402 list_remove(&iface->port_elem);
3403 hmap_remove(&br->iface_by_name, &iface->name_node);
3405 netdev_close(iface->netdev);
3412 static struct iface *
3413 iface_lookup(const struct bridge *br, const char *name)
3415 struct iface *iface;
3417 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3418 &br->iface_by_name) {
3419 if (!strcmp(iface->name, name)) {
3427 static struct iface *
3428 iface_find(const char *name)
3430 const struct bridge *br;
3432 HMAP_FOR_EACH (br, node, &all_bridges) {
3433 struct iface *iface = iface_lookup(br, name);
3442 static struct if_cfg *
3443 if_cfg_lookup(const struct bridge *br, const char *name)
3445 struct if_cfg *if_cfg;
3447 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3449 if (!strcmp(if_cfg->cfg->name, name)) {
3457 static struct iface *
3458 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3460 struct iface *iface;
3462 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3463 hash_int(ofp_port, 0), &br->ifaces) {
3464 if (iface->ofp_port == ofp_port) {
3471 /* Set Ethernet address of 'iface', if one is specified in the configuration
3474 iface_set_mac(struct iface *iface)
3476 uint8_t ea[ETH_ADDR_LEN];
3478 if (!strcmp(iface->type, "internal")
3479 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3480 if (iface->ofp_port == OFPP_LOCAL) {
3481 VLOG_ERR("interface %s: ignoring mac in Interface record "
3482 "(use Bridge record to set local port's mac)",
3484 } else if (eth_addr_is_multicast(ea)) {
3485 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3488 int error = netdev_set_etheraddr(iface->netdev, ea);
3490 VLOG_ERR("interface %s: setting MAC failed (%s)",
3491 iface->name, strerror(error));
3497 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3499 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3501 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3502 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3506 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3507 * sets the "ofport" field to -1.
3509 * This is appropriate when 'if_cfg''s interface cannot be created or is
3510 * otherwise invalid. */
3512 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3514 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3515 ovsrec_interface_set_status(if_cfg, NULL);
3516 ovsrec_interface_set_admin_state(if_cfg, NULL);
3517 ovsrec_interface_set_duplex(if_cfg, NULL);
3518 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3519 ovsrec_interface_set_link_state(if_cfg, NULL);
3520 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3521 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3522 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3523 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3524 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3525 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3526 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3530 struct iface_delete_queues_cbdata {
3531 struct netdev *netdev;
3532 const struct ovsdb_datum *queues;
3536 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3538 union ovsdb_atom atom;
3540 atom.integer = target;
3541 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3545 iface_delete_queues(unsigned int queue_id,
3546 const struct smap *details OVS_UNUSED, void *cbdata_)
3548 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3550 if (!queue_ids_include(cbdata->queues, queue_id)) {
3551 netdev_delete_queue(cbdata->netdev, queue_id);
3556 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3558 struct ofpbuf queues_buf;
3560 ofpbuf_init(&queues_buf, 0);
3562 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3563 netdev_set_qos(iface->netdev, NULL, NULL);
3565 struct iface_delete_queues_cbdata cbdata;
3569 /* Configure top-level Qos for 'iface'. */
3570 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3572 /* Deconfigure queues that were deleted. */
3573 cbdata.netdev = iface->netdev;
3574 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3576 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3578 /* Configure queues for 'iface'. */
3580 for (i = 0; i < qos->n_queues; i++) {
3581 const struct ovsrec_queue *queue = qos->value_queues[i];
3582 unsigned int queue_id = qos->key_queues[i];
3584 if (queue_id == 0) {
3588 if (queue->n_dscp == 1) {
3589 struct ofproto_port_queue *port_queue;
3591 port_queue = ofpbuf_put_uninit(&queues_buf,
3592 sizeof *port_queue);
3593 port_queue->queue = queue_id;
3594 port_queue->dscp = queue->dscp[0];
3597 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3600 struct smap details;
3602 smap_init(&details);
3603 netdev_set_queue(iface->netdev, 0, &details);
3604 smap_destroy(&details);
3608 if (iface->ofp_port >= 0) {
3609 const struct ofproto_port_queue *port_queues = queues_buf.data;
3610 size_t n_queues = queues_buf.size / sizeof *port_queues;
3612 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3613 port_queues, n_queues);
3616 netdev_set_policing(iface->netdev,
3617 iface->cfg->ingress_policing_rate,
3618 iface->cfg->ingress_policing_burst);
3620 ofpbuf_uninit(&queues_buf);
3624 iface_configure_cfm(struct iface *iface)
3626 const struct ovsrec_interface *cfg = iface->cfg;
3627 const char *opstate_str;
3628 const char *cfm_ccm_vlan;
3629 struct cfm_settings s;
3630 struct smap netdev_args;
3632 if (!cfg->n_cfm_mpid) {
3633 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3637 s.check_tnl_key = false;
3638 smap_init(&netdev_args);
3639 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3640 const char *key = smap_get(&netdev_args, "key");
3641 const char *in_key = smap_get(&netdev_args, "in_key");
3643 s.check_tnl_key = (key && !strcmp(key, "flow"))
3644 || (in_key && !strcmp(in_key, "flow"));
3646 smap_destroy(&netdev_args);
3648 s.mpid = *cfg->cfm_mpid;
3649 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3650 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3651 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3653 if (s.interval <= 0) {
3657 if (!cfm_ccm_vlan) {
3659 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3660 s.ccm_vlan = CFM_RANDOM_VLAN;
3662 s.ccm_vlan = atoi(cfm_ccm_vlan);
3663 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3668 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3671 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3672 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3674 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3677 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3678 * instead of obtaining it from the database. */
3680 iface_is_synthetic(const struct iface *iface)
3682 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3686 iface_pick_ofport(const struct ovsrec_interface *cfg)
3688 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3689 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3693 /* Port mirroring. */
3695 static struct mirror *
3696 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3700 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3701 if (uuid_equals(uuid, &m->uuid)) {
3709 bridge_configure_mirrors(struct bridge *br)
3711 const struct ovsdb_datum *mc;
3712 unsigned long *flood_vlans;
3713 struct mirror *m, *next;
3716 /* Get rid of deleted mirrors. */
3717 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3718 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3719 union ovsdb_atom atom;
3721 atom.uuid = m->uuid;
3722 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3727 /* Add new mirrors and reconfigure existing ones. */
3728 for (i = 0; i < br->cfg->n_mirrors; i++) {
3729 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3730 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3732 m = mirror_create(br, cfg);
3735 if (!mirror_configure(m)) {
3740 /* Update flooded vlans (for RSPAN). */
3741 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3742 br->cfg->n_flood_vlans);
3743 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3744 bitmap_free(flood_vlans);
3747 static struct mirror *
3748 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3752 m = xzalloc(sizeof *m);
3753 m->uuid = cfg->header_.uuid;
3754 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3756 m->name = xstrdup(cfg->name);
3762 mirror_destroy(struct mirror *m)
3765 struct bridge *br = m->bridge;
3768 ofproto_mirror_unregister(br->ofproto, m);
3771 hmap_remove(&br->mirrors, &m->hmap_node);
3778 mirror_collect_ports(struct mirror *m,
3779 struct ovsrec_port **in_ports, int n_in_ports,
3780 void ***out_portsp, size_t *n_out_portsp)
3782 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3783 size_t n_out_ports = 0;
3786 for (i = 0; i < n_in_ports; i++) {
3787 const char *name = in_ports[i]->name;
3788 struct port *port = port_lookup(m->bridge, name);
3790 out_ports[n_out_ports++] = port;
3792 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3793 "port %s", m->bridge->name, m->name, name);
3796 *out_portsp = out_ports;
3797 *n_out_portsp = n_out_ports;
3801 mirror_configure(struct mirror *m)
3803 const struct ovsrec_mirror *cfg = m->cfg;
3804 struct ofproto_mirror_settings s;
3807 if (strcmp(cfg->name, m->name)) {
3809 m->name = xstrdup(cfg->name);
3813 /* Get output port or VLAN. */
3814 if (cfg->output_port) {
3815 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3816 if (!s.out_bundle) {
3817 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3818 m->bridge->name, m->name);
3821 s.out_vlan = UINT16_MAX;
3823 if (cfg->output_vlan) {
3824 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3825 "output vlan; ignoring output vlan",
3826 m->bridge->name, m->name);
3828 } else if (cfg->output_vlan) {
3829 /* The database should prevent invalid VLAN values. */
3830 s.out_bundle = NULL;
3831 s.out_vlan = *cfg->output_vlan;
3833 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3834 m->bridge->name, m->name);
3838 /* Get port selection. */
3839 if (cfg->select_all) {
3840 size_t n_ports = hmap_count(&m->bridge->ports);
3841 void **ports = xmalloc(n_ports * sizeof *ports);
3846 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3856 /* Get ports, dropping ports that don't exist.
3857 * The IDL ensures that there are no duplicates. */
3858 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3859 &s.srcs, &s.n_srcs);
3860 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3861 &s.dsts, &s.n_dsts);
3864 /* Get VLAN selection. */
3865 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3868 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3871 if (s.srcs != s.dsts) {
3880 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3882 * This is deprecated. It is only for compatibility with broken device drivers
3883 * in old versions of Linux that do not properly support VLANs when VLAN
3884 * devices are not used. When broken device drivers are no longer in
3885 * widespread use, we will delete these interfaces. */
3887 static struct ovsrec_port **recs;
3888 static size_t n_recs, allocated_recs;
3890 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3891 * splinters are reconfigured. */
3893 register_rec(struct ovsrec_port *rec)
3895 if (n_recs >= allocated_recs) {
3896 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3898 recs[n_recs++] = rec;
3901 /* Frees all of the ports registered with register_reg(). */
3903 free_registered_recs(void)
3907 for (i = 0; i < n_recs; i++) {
3908 struct ovsrec_port *port = recs[i];
3911 for (j = 0; j < port->n_interfaces; j++) {
3912 struct ovsrec_interface *iface = port->interfaces[j];
3917 smap_destroy(&port->other_config);
3918 free(port->interfaces);
3926 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3929 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3931 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3935 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3938 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3939 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3940 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3943 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3944 * use, returns NULL.
3946 * Updates 'vlan_splinters_enabled_anywhere'. */
3947 static unsigned long int *
3948 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3950 unsigned long int *splinter_vlans;
3951 struct sset splinter_ifaces;
3952 const char *real_dev_name;
3953 struct shash *real_devs;
3954 struct shash_node *node;
3958 /* Free space allocated for synthesized ports and interfaces, since we're
3959 * in the process of reconstructing all of them. */
3960 free_registered_recs();
3962 splinter_vlans = bitmap_allocate(4096);
3963 sset_init(&splinter_ifaces);
3964 vlan_splinters_enabled_anywhere = false;
3965 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3966 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3969 for (j = 0; j < br_cfg->n_ports; j++) {
3970 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3973 for (k = 0; k < port_cfg->n_interfaces; k++) {
3974 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3976 if (vlan_splinters_is_enabled(iface_cfg)) {
3977 vlan_splinters_enabled_anywhere = true;
3978 sset_add(&splinter_ifaces, iface_cfg->name);
3979 vlan_bitmap_from_array__(port_cfg->trunks,
3985 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3986 bitmap_set1(splinter_vlans, *port_cfg->tag);
3991 if (!vlan_splinters_enabled_anywhere) {
3992 free(splinter_vlans);
3993 sset_destroy(&splinter_ifaces);
3997 HMAP_FOR_EACH (br, node, &all_bridges) {
3999 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4003 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4004 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4005 * device to be created for it. */
4006 bitmap_set0(splinter_vlans, 0);
4007 bitmap_set0(splinter_vlans, 4095);
4009 /* Delete all VLAN devices that we don't need. */
4011 real_devs = vlandev_get_real_devs();
4012 SHASH_FOR_EACH (node, real_devs) {
4013 const struct vlan_real_dev *real_dev = node->data;
4014 const struct vlan_dev *vlan_dev;
4015 bool real_dev_has_splinters;
4017 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4019 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4020 if (!real_dev_has_splinters
4021 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4022 struct netdev *netdev;
4024 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4025 if (!netdev_get_in4(netdev, NULL, NULL) ||
4026 !netdev_get_in6(netdev, NULL)) {
4027 vlandev_del(vlan_dev->name);
4029 /* It has an IP address configured, so we don't own
4030 * it. Don't delete it. */
4032 netdev_close(netdev);
4039 /* Add all VLAN devices that we need. */
4040 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4043 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4044 if (!vlandev_get_name(real_dev_name, vid)) {
4045 vlandev_add(real_dev_name, vid);
4052 sset_destroy(&splinter_ifaces);
4054 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4055 free(splinter_vlans);
4058 return splinter_vlans;
4061 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4064 configure_splinter_port(struct port *port)
4066 struct ofproto *ofproto = port->bridge->ofproto;
4067 uint16_t realdev_ofp_port;
4068 const char *realdev_name;
4069 struct iface *vlandev, *realdev;
4071 ofproto_bundle_unregister(port->bridge->ofproto, port);
4073 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4076 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4077 realdev = iface_lookup(port->bridge, realdev_name);
4078 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4080 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4084 static struct ovsrec_port *
4085 synthesize_splinter_port(const char *real_dev_name,
4086 const char *vlan_dev_name, int vid)
4088 struct ovsrec_interface *iface;
4089 struct ovsrec_port *port;
4091 iface = xmalloc(sizeof *iface);
4092 ovsrec_interface_init(iface);
4093 iface->name = xstrdup(vlan_dev_name);
4094 iface->type = "system";
4096 port = xmalloc(sizeof *port);
4097 ovsrec_port_init(port);
4098 port->interfaces = xmemdup(&iface, sizeof iface);
4099 port->n_interfaces = 1;
4100 port->name = xstrdup(vlan_dev_name);
4101 port->vlan_mode = "splinter";
4102 port->tag = xmalloc(sizeof *port->tag);
4105 smap_add(&port->other_config, "realdev", real_dev_name);
4111 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4112 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4113 * 1-bit in the 'splinter_vlans' bitmap. */
4115 add_vlan_splinter_ports(struct bridge *br,
4116 const unsigned long int *splinter_vlans,
4117 struct shash *ports)
4121 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4122 * we're modifying 'ports'. */
4123 for (i = 0; i < br->cfg->n_ports; i++) {
4124 const char *name = br->cfg->ports[i]->name;
4125 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4128 for (j = 0; j < port_cfg->n_interfaces; j++) {
4129 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4131 if (vlan_splinters_is_enabled(iface_cfg)) {
4132 const char *real_dev_name;
4135 real_dev_name = iface_cfg->name;
4136 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4137 const char *vlan_dev_name;
4139 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4141 && !shash_find(ports, vlan_dev_name)) {
4142 shash_add(ports, vlan_dev_name,
4143 synthesize_splinter_port(
4144 real_dev_name, vlan_dev_name, vid));
4153 mirror_refresh_stats(struct mirror *m)
4155 struct ofproto *ofproto = m->bridge->ofproto;
4156 uint64_t tx_packets, tx_bytes;
4159 size_t stat_cnt = 0;
4161 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4162 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4166 if (tx_packets != UINT64_MAX) {
4167 keys[stat_cnt] = "tx_packets";
4168 values[stat_cnt] = tx_packets;
4171 if (tx_bytes != UINT64_MAX) {
4172 keys[stat_cnt] = "tx_bytes";
4173 values[stat_cnt] = tx_bytes;
4177 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);