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_stp(struct bridge *);
204 static void bridge_configure_tables(struct bridge *);
205 static void bridge_configure_dp_desc(struct bridge *);
206 static void bridge_configure_remotes(struct bridge *,
207 const struct sockaddr_in *managers,
209 static void bridge_pick_local_hw_addr(struct bridge *,
210 uint8_t ea[ETH_ADDR_LEN],
211 struct iface **hw_addr_iface);
212 static uint64_t bridge_pick_datapath_id(struct bridge *,
213 const uint8_t bridge_ea[ETH_ADDR_LEN],
214 struct iface *hw_addr_iface);
215 static void bridge_queue_if_cfg(struct bridge *,
216 const struct ovsrec_interface *,
217 const struct ovsrec_port *);
218 static uint64_t dpid_from_hash(const void *, size_t nbytes);
219 static bool bridge_has_bond_fake_iface(const struct bridge *,
221 static bool port_is_bond_fake_iface(const struct port *);
223 static unixctl_cb_func qos_unixctl_show;
225 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
226 static void port_del_ifaces(struct port *);
227 static void port_destroy(struct port *);
228 static struct port *port_lookup(const struct bridge *, const char *name);
229 static void port_configure(struct port *);
230 static struct lacp_settings *port_configure_lacp(struct port *,
231 struct lacp_settings *);
232 static void port_configure_bond(struct port *, struct bond_settings *);
233 static bool port_is_synthetic(const struct port *);
235 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
236 static void run_system_stats(void);
238 static void bridge_configure_mirrors(struct bridge *);
239 static struct mirror *mirror_create(struct bridge *,
240 const struct ovsrec_mirror *);
241 static void mirror_destroy(struct mirror *);
242 static bool mirror_configure(struct mirror *);
243 static void mirror_refresh_stats(struct mirror *);
245 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
246 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
247 static bool iface_is_internal(const struct ovsrec_interface *iface,
248 const struct ovsrec_bridge *br);
249 static const char *iface_get_type(const struct ovsrec_interface *,
250 const struct ovsrec_bridge *);
251 static void iface_destroy(struct iface *);
252 static struct iface *iface_lookup(const struct bridge *, const char *name);
253 static struct iface *iface_find(const char *name);
254 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
255 static struct iface *iface_from_ofp_port(const struct bridge *,
257 static void iface_set_mac(struct iface *);
258 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
259 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
260 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
261 static void iface_configure_cfm(struct iface *);
262 static void iface_refresh_cfm_stats(struct iface *);
263 static void iface_refresh_stats(struct iface *);
264 static void iface_refresh_status(struct iface *);
265 static bool iface_is_synthetic(const struct iface *);
266 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
268 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
270 * This is deprecated. It is only for compatibility with broken device drivers
271 * in old versions of Linux that do not properly support VLANs when VLAN
272 * devices are not used. When broken device drivers are no longer in
273 * widespread use, we will delete these interfaces. */
275 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
276 static bool vlan_splinters_enabled_anywhere;
278 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
279 static unsigned long int *collect_splinter_vlans(
280 const struct ovsrec_open_vswitch *);
281 static void configure_splinter_port(struct port *);
282 static void add_vlan_splinter_ports(struct bridge *,
283 const unsigned long int *splinter_vlans,
284 struct shash *ports);
287 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
289 struct shash iface_hints;
290 static bool initialized = false;
297 shash_init(&iface_hints);
300 for (i = 0; i < cfg->n_bridges; i++) {
301 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
304 for (j = 0; j < br_cfg->n_ports; j++) {
305 struct ovsrec_port *port_cfg = br_cfg->ports[j];
308 for (k = 0; k < port_cfg->n_interfaces; k++) {
309 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
310 struct iface_hint *iface_hint;
312 iface_hint = xmalloc(sizeof *iface_hint);
313 iface_hint->br_name = br_cfg->name;
314 iface_hint->br_type = br_cfg->datapath_type;
315 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
317 shash_add(&iface_hints, if_cfg->name, iface_hint);
323 ofproto_init(&iface_hints);
325 shash_destroy_free_data(&iface_hints);
329 /* Public functions. */
331 /* Initializes the bridge module, configuring it to obtain its configuration
332 * from an OVSDB server accessed over 'remote', which should be a string in a
333 * form acceptable to ovsdb_idl_create(). */
335 bridge_init(const char *remote)
337 /* Create connection to database. */
338 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
339 idl_seqno = ovsdb_idl_get_seqno(idl);
340 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
341 ovsdb_idl_verify_write_only(idl);
343 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
345 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
351 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
353 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
355 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
357 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
358 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
360 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
376 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
378 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
379 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
381 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
383 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
385 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
388 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
390 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
396 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
400 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
402 /* Register unixctl commands. */
403 unixctl_command_register("qos/show", "interface", 1, 1,
404 qos_unixctl_show, NULL);
405 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
406 bridge_unixctl_dump_flows, NULL);
407 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
408 bridge_unixctl_reconnect, NULL);
418 struct bridge *br, *next_br;
420 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
423 ovsdb_idl_destroy(idl);
426 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
427 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
428 * responsible for freeing '*managersp' (with free()).
430 * You may be asking yourself "why does ovs-vswitchd care?", because
431 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
432 * should not be and in fact is not directly involved in that. But
433 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
434 * it has to tell in-band control where the managers are to enable that.
435 * (Thus, only managers connected in-band are collected.)
438 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
439 struct sockaddr_in **managersp, size_t *n_managersp)
441 struct sockaddr_in *managers = NULL;
442 size_t n_managers = 0;
446 /* Collect all of the potential targets from the "targets" columns of the
447 * rows pointed to by "manager_options", excluding any that are
450 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
451 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
453 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
454 sset_find_and_delete(&targets, m->target);
456 sset_add(&targets, m->target);
460 /* Now extract the targets' IP addresses. */
461 if (!sset_is_empty(&targets)) {
464 managers = xmalloc(sset_count(&targets) * sizeof *managers);
465 SSET_FOR_EACH (target, &targets) {
466 struct sockaddr_in *sin = &managers[n_managers];
468 if (stream_parse_target_with_default_ports(target,
476 sset_destroy(&targets);
478 *managersp = managers;
479 *n_managersp = n_managers;
483 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
485 unsigned long int *splinter_vlans;
488 COVERAGE_INC(bridge_reconfigure);
490 ovs_assert(!reconfiguring);
491 reconfiguring = true;
493 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
494 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
495 * configuration otherwise.
497 * This is mostly an update to bridge data structures. Nothing is pushed
498 * down to ofproto or lower layers. */
499 add_del_bridges(ovs_cfg);
500 splinter_vlans = collect_splinter_vlans(ovs_cfg);
501 HMAP_FOR_EACH (br, node, &all_bridges) {
502 bridge_add_del_ports(br, splinter_vlans);
504 free(splinter_vlans);
506 /* Delete datapaths that are no longer configured, and create ones which
507 * don't exist but should. */
508 bridge_update_ofprotos();
510 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
511 HMAP_FOR_EACH (br, node, &all_bridges) {
512 bridge_refresh_ofp_port(br);
515 /* Clear database records for "if_cfg"s which haven't been instantiated. */
516 HMAP_FOR_EACH (br, node, &all_bridges) {
517 struct if_cfg *if_cfg;
519 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
520 iface_clear_db_record(if_cfg->cfg);
524 reconfigure_system_stats(ovs_cfg);
528 bridge_reconfigure_ofp(void)
530 long long int deadline;
534 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
536 /* The kernel will reject any attempt to add a given port to a datapath if
537 * that port already belongs to a different datapath, so we must do all
538 * port deletions before any port additions. */
539 HMAP_FOR_EACH (br, node, &all_bridges) {
540 struct ofpp_garbage *garbage, *next;
542 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
543 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
544 * internal datastructures may not be consistent. Eventually, when
545 * port additions and deletions are cheaper, these calls should be
548 ofproto_port_del(br->ofproto, garbage->ofp_port);
549 list_remove(&garbage->list_node);
553 if (time_msec() >= deadline) {
560 HMAP_FOR_EACH (br, node, &all_bridges) {
561 struct if_cfg *if_cfg, *next;
563 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
564 iface_create(br, if_cfg, -1);
566 if (time_msec() >= deadline) {
576 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
578 struct sockaddr_in *managers;
579 int sflow_bridge_number;
584 ovs_assert(reconfiguring);
585 done = bridge_reconfigure_ofp();
587 /* Complete the configuration. */
588 sflow_bridge_number = 0;
589 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
590 HMAP_FOR_EACH (br, node, &all_bridges) {
593 /* We need the datapath ID early to allow LACP ports to use it as the
594 * default system ID. */
595 bridge_configure_datapath_id(br);
597 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
600 port_configure(port);
602 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
603 iface_configure_cfm(iface);
604 iface_configure_qos(iface, port->cfg->qos);
605 iface_set_mac(iface);
608 bridge_configure_mirrors(br);
609 bridge_configure_flow_eviction_threshold(br);
610 bridge_configure_forward_bpdu(br);
611 bridge_configure_mac_table(br);
612 bridge_configure_remotes(br, managers, n_managers);
613 bridge_configure_netflow(br);
614 bridge_configure_sflow(br, &sflow_bridge_number);
615 bridge_configure_stp(br);
616 bridge_configure_tables(br);
617 bridge_configure_dp_desc(br);
624 /* Delete ofprotos which aren't configured or have the wrong type. Create
625 * ofprotos which don't exist but need to. */
627 bridge_update_ofprotos(void)
629 struct bridge *br, *next;
634 /* Delete ofprotos with no bridge or with the wrong type. */
637 ofproto_enumerate_types(&types);
638 SSET_FOR_EACH (type, &types) {
641 ofproto_enumerate_names(type, &names);
642 SSET_FOR_EACH (name, &names) {
643 br = bridge_lookup(name);
644 if (!br || strcmp(type, br->type)) {
645 ofproto_delete(name, type);
649 sset_destroy(&names);
650 sset_destroy(&types);
652 /* Add ofprotos for bridges which don't have one yet. */
653 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
661 /* Remove ports from any datapath with the same name as 'br'. If we
662 * don't do this, creating 'br''s ofproto will fail because a port with
663 * the same name as its local port already exists. */
664 HMAP_FOR_EACH (br2, node, &all_bridges) {
665 struct ofproto_port ofproto_port;
671 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
673 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
675 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
678 ofproto_port_destroy(&ofproto_port);
682 error = ofproto_create(br->name, br->type, &br->ofproto);
684 VLOG_ERR("failed to create bridge %s: %s", br->name,
692 port_configure(struct port *port)
694 const struct ovsrec_port *cfg = port->cfg;
695 struct bond_settings bond_settings;
696 struct lacp_settings lacp_settings;
697 struct ofproto_bundle_settings s;
700 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
701 configure_splinter_port(port);
710 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
711 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
712 s.slaves[s.n_slaves++] = iface->ofp_port;
717 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
721 /* Get VLAN trunks. */
724 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
728 if (cfg->vlan_mode) {
729 if (!strcmp(cfg->vlan_mode, "access")) {
730 s.vlan_mode = PORT_VLAN_ACCESS;
731 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
732 s.vlan_mode = PORT_VLAN_TRUNK;
733 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
734 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
735 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
736 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
738 /* This "can't happen" because ovsdb-server should prevent it. */
739 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
740 s.vlan_mode = PORT_VLAN_TRUNK;
744 s.vlan_mode = PORT_VLAN_ACCESS;
746 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
750 s.vlan_mode = PORT_VLAN_TRUNK;
753 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
756 /* Get LACP settings. */
757 s.lacp = port_configure_lacp(port, &lacp_settings);
761 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
762 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
763 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
766 s.lacp_slaves = NULL;
769 /* Get bond settings. */
770 if (s.n_slaves > 1) {
771 s.bond = &bond_settings;
772 port_configure_bond(port, &bond_settings);
775 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
776 netdev_set_miimon_interval(iface->netdev, 0);
781 ofproto_bundle_register(port->bridge->ofproto, port, &s);
789 /* Pick local port hardware address and datapath ID for 'br'. */
791 bridge_configure_datapath_id(struct bridge *br)
793 uint8_t ea[ETH_ADDR_LEN];
795 struct iface *local_iface;
796 struct iface *hw_addr_iface;
799 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
800 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
802 int error = netdev_set_etheraddr(local_iface->netdev, ea);
804 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
805 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
806 "Ethernet address: %s",
807 br->name, strerror(error));
810 memcpy(br->ea, ea, ETH_ADDR_LEN);
812 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
813 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
814 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
815 ofproto_set_datapath_id(br->ofproto, dpid);
818 dpid_string = xasprintf("%016"PRIx64, dpid);
819 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
823 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
826 bridge_get_allowed_versions(struct bridge *br)
828 if (!br->cfg->n_protocols)
831 return ofputil_versions_from_strings(br->cfg->protocols,
832 br->cfg->n_protocols);
835 /* Set NetFlow configuration on 'br'. */
837 bridge_configure_netflow(struct bridge *br)
839 struct ovsrec_netflow *cfg = br->cfg->netflow;
840 struct netflow_options opts;
843 ofproto_set_netflow(br->ofproto, NULL);
847 memset(&opts, 0, sizeof opts);
849 /* Get default NetFlow configuration from datapath.
850 * Apply overrides from 'cfg'. */
851 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
852 if (cfg->engine_type) {
853 opts.engine_type = *cfg->engine_type;
855 if (cfg->engine_id) {
856 opts.engine_id = *cfg->engine_id;
859 /* Configure active timeout interval. */
860 opts.active_timeout = cfg->active_timeout;
861 if (!opts.active_timeout) {
862 opts.active_timeout = -1;
863 } else if (opts.active_timeout < 0) {
864 VLOG_WARN("bridge %s: active timeout interval set to negative "
865 "value, using default instead (%d seconds)", br->name,
866 NF_ACTIVE_TIMEOUT_DEFAULT);
867 opts.active_timeout = -1;
870 /* Add engine ID to interface number to disambiguate bridgs? */
871 opts.add_id_to_iface = cfg->add_id_to_interface;
872 if (opts.add_id_to_iface) {
873 if (opts.engine_id > 0x7f) {
874 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
875 "another vswitch, choose an engine id less than 128",
878 if (hmap_count(&br->ports) > 508) {
879 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
880 "another port when more than 508 ports are used",
886 sset_init(&opts.collectors);
887 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
890 if (ofproto_set_netflow(br->ofproto, &opts)) {
891 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
893 sset_destroy(&opts.collectors);
896 /* Set sFlow configuration on 'br'. */
898 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
900 const struct ovsrec_sflow *cfg = br->cfg->sflow;
901 struct ovsrec_controller **controllers;
902 struct ofproto_sflow_options oso;
903 size_t n_controllers;
907 ofproto_set_sflow(br->ofproto, NULL);
911 memset(&oso, 0, sizeof oso);
913 sset_init(&oso.targets);
914 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
916 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
918 oso.sampling_rate = *cfg->sampling;
921 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
923 oso.polling_interval = *cfg->polling;
926 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
928 oso.header_len = *cfg->header;
931 oso.sub_id = (*sflow_bridge_number)++;
932 oso.agent_device = cfg->agent;
934 oso.control_ip = NULL;
935 n_controllers = bridge_get_controllers(br, &controllers);
936 for (i = 0; i < n_controllers; i++) {
937 if (controllers[i]->local_ip) {
938 oso.control_ip = controllers[i]->local_ip;
942 ofproto_set_sflow(br->ofproto, &oso);
944 sset_destroy(&oso.targets);
948 port_configure_stp(const struct ofproto *ofproto, struct port *port,
949 struct ofproto_port_stp_settings *port_s,
950 int *port_num_counter, unsigned long *port_num_bitmap)
952 const char *config_str;
955 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
956 port_s->enable = false;
959 port_s->enable = true;
962 /* STP over bonds is not supported. */
963 if (!list_is_singleton(&port->ifaces)) {
964 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
966 port_s->enable = false;
970 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
972 /* Internal ports shouldn't participate in spanning tree, so
974 if (!strcmp(iface->type, "internal")) {
975 VLOG_DBG("port %s: disable STP on internal ports", port->name);
976 port_s->enable = false;
980 /* STP on mirror output ports is not supported. */
981 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
982 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
983 port_s->enable = false;
987 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
989 unsigned long int port_num = strtoul(config_str, NULL, 0);
990 int port_idx = port_num - 1;
992 if (port_num < 1 || port_num > STP_MAX_PORTS) {
993 VLOG_ERR("port %s: invalid stp-port-num", port->name);
994 port_s->enable = false;
998 if (bitmap_is_set(port_num_bitmap, port_idx)) {
999 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1000 port->name, port_num);
1001 port_s->enable = false;
1004 bitmap_set1(port_num_bitmap, port_idx);
1005 port_s->port_num = port_idx;
1007 if (*port_num_counter >= STP_MAX_PORTS) {
1008 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1009 port_s->enable = false;
1013 port_s->port_num = (*port_num_counter)++;
1016 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1018 port_s->path_cost = strtoul(config_str, NULL, 10);
1020 enum netdev_features current;
1023 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1024 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1025 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1028 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1030 port_s->priority = strtoul(config_str, NULL, 0);
1032 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1036 /* Set spanning tree configuration on 'br'. */
1038 bridge_configure_stp(struct bridge *br)
1040 if (!br->cfg->stp_enable) {
1041 ofproto_set_stp(br->ofproto, NULL);
1043 struct ofproto_stp_settings br_s;
1044 const char *config_str;
1046 int port_num_counter;
1047 unsigned long *port_num_bitmap;
1049 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1051 uint8_t ea[ETH_ADDR_LEN];
1053 if (eth_addr_from_string(config_str, ea)) {
1054 br_s.system_id = eth_addr_to_uint64(ea);
1056 br_s.system_id = eth_addr_to_uint64(br->ea);
1057 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1058 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1061 br_s.system_id = eth_addr_to_uint64(br->ea);
1064 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1066 br_s.priority = strtoul(config_str, NULL, 0);
1068 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1071 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1073 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1075 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1078 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1080 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1082 br_s.max_age = STP_DEFAULT_MAX_AGE;
1085 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1087 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1089 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1092 /* Configure STP on the bridge. */
1093 if (ofproto_set_stp(br->ofproto, &br_s)) {
1094 VLOG_ERR("bridge %s: could not enable STP", br->name);
1098 /* Users must either set the port number with the "stp-port-num"
1099 * configuration on all ports or none. If manual configuration
1100 * is not done, then we allocate them sequentially. */
1101 port_num_counter = 0;
1102 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1103 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1104 struct ofproto_port_stp_settings port_s;
1105 struct iface *iface;
1107 port_configure_stp(br->ofproto, port, &port_s,
1108 &port_num_counter, port_num_bitmap);
1110 /* As bonds are not supported, just apply configuration to
1111 * all interfaces. */
1112 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1113 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1115 VLOG_ERR("port %s: could not enable STP", port->name);
1121 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1122 && port_num_counter) {
1123 VLOG_ERR("bridge %s: must manually configure all STP port "
1124 "IDs or none, disabling", br->name);
1125 ofproto_set_stp(br->ofproto, NULL);
1127 bitmap_free(port_num_bitmap);
1132 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1134 const struct port *port = port_lookup(br, name);
1135 return port && port_is_bond_fake_iface(port);
1139 port_is_bond_fake_iface(const struct port *port)
1141 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1145 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1147 struct bridge *br, *next;
1148 struct shash new_br;
1151 /* Collect new bridges' names and types. */
1152 shash_init(&new_br);
1153 for (i = 0; i < cfg->n_bridges; i++) {
1154 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1155 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1157 if (strchr(br_cfg->name, '/')) {
1158 /* Prevent remote ovsdb-server users from accessing arbitrary
1159 * directories, e.g. consider a bridge named "../../../etc/". */
1160 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1162 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1163 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1167 /* Get rid of deleted bridges or those whose types have changed.
1168 * Update 'cfg' of bridges that still exist. */
1169 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1170 br->cfg = shash_find_data(&new_br, br->name);
1171 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1172 br->cfg->datapath_type))) {
1177 /* Add new bridges. */
1178 for (i = 0; i < cfg->n_bridges; i++) {
1179 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1180 struct bridge *br = bridge_lookup(br_cfg->name);
1182 bridge_create(br_cfg);
1186 shash_destroy(&new_br);
1190 iface_set_ofp_port(struct iface *iface, int ofp_port)
1192 struct bridge *br = iface->port->bridge;
1194 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1195 iface->ofp_port = ofp_port;
1196 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1197 iface_set_ofport(iface->cfg, ofp_port);
1200 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1201 * Returns 0 if successful, otherwise a positive errno value. */
1203 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1204 struct netdev *netdev)
1208 error = netdev_set_config(netdev, &iface_cfg->options);
1210 VLOG_WARN("could not configure network device %s (%s)",
1211 iface_cfg->name, strerror(error));
1216 /* This function determines whether 'ofproto_port', which is attached to
1217 * br->ofproto's datapath, is one that we want in 'br'.
1219 * If it is, it returns true, after creating an iface (if necessary),
1220 * configuring the iface's netdev according to the iface's options, and setting
1221 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1223 * If, on the other hand, 'port' should be removed, it returns false. The
1224 * caller should later detach the port from br->ofproto. */
1226 bridge_refresh_one_ofp_port(struct bridge *br,
1227 const struct ofproto_port *ofproto_port)
1229 const char *name = ofproto_port->name;
1230 const char *type = ofproto_port->type;
1231 uint16_t ofp_port = ofproto_port->ofp_port;
1233 struct iface *iface = iface_lookup(br, name);
1235 /* Check that the name-to-number mapping is one-to-one. */
1236 if (iface->ofp_port >= 0) {
1237 VLOG_WARN("bridge %s: interface %s reported twice",
1240 } else if (iface_from_ofp_port(br, ofp_port)) {
1241 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1242 br->name, ofp_port);
1246 /* There's a configured interface named 'name'. */
1247 if (strcmp(type, iface->type)
1248 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1249 /* It's the wrong type, or it's the right type but can't be
1250 * configured as the user requested, so we must destroy it. */
1253 /* It's the right type and configured correctly. Keep it. */
1254 iface_set_ofp_port(iface, ofp_port);
1257 } else if (bridge_has_bond_fake_iface(br, name)
1258 && !strcmp(type, "internal")) {
1259 /* It's a bond fake iface. Keep it. */
1262 /* There's no configured interface named 'name', but there might be an
1263 * interface of that name queued to be created.
1265 * If there is, and it has the correct type, then try to configure it
1266 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1267 * delete it and later try to re-add it. */
1268 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1270 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1271 && iface_create(br, if_cfg, ofp_port));
1275 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1276 * consistent with the ofp_ports in "br->ofproto". */
1278 bridge_refresh_ofp_port(struct bridge *br)
1280 struct ofproto_port_dump dump;
1281 struct ofproto_port ofproto_port;
1282 struct port *port, *port_next;
1284 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1285 hmap_clear(&br->ifaces);
1286 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1287 struct iface *iface;
1289 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1290 iface->ofp_port = -1;
1294 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1295 * already exist in the datapath and promote them to full fledged "struct
1296 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1297 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1298 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1299 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1300 garbage->ofp_port = ofproto_port.ofp_port;
1301 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1305 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1306 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1307 * later they can be added to ofproto. */
1308 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1309 struct iface *iface, *iface_next;
1311 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1312 if (iface->ofp_port < 0) {
1313 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1314 iface_destroy(iface);
1318 if (list_is_empty(&port->ifaces)) {
1324 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1325 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1326 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1329 * If successful, returns 0 and stores the network device in '*netdevp'. On
1330 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1332 iface_do_create(const struct bridge *br,
1333 const struct if_cfg *if_cfg,
1334 int *ofp_portp, struct netdev **netdevp)
1336 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1337 const struct ovsrec_port *port_cfg = if_cfg->parent;
1338 struct netdev *netdev;
1341 error = netdev_open(iface_cfg->name,
1342 iface_get_type(iface_cfg, br->cfg), &netdev);
1344 VLOG_WARN("could not open network device %s (%s)",
1345 iface_cfg->name, strerror(error));
1349 error = iface_set_netdev_config(iface_cfg, netdev);
1354 if (*ofp_portp < 0) {
1355 uint16_t ofp_port = if_cfg->ofport;
1357 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1361 *ofp_portp = ofp_port;
1363 VLOG_INFO("bridge %s: added interface %s on port %d",
1364 br->name, iface_cfg->name, *ofp_portp);
1366 VLOG_DBG("bridge %s: interface %s is on port %d",
1367 br->name, iface_cfg->name, *ofp_portp);
1370 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1371 || iface_is_internal(iface_cfg, br->cfg)) {
1372 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1380 netdev_close(netdev);
1384 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1385 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1386 * automatically allocated for the iface. Takes ownership of and
1387 * deallocates 'if_cfg'.
1389 * Return true if an iface is successfully created, false otherwise. */
1391 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1393 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1394 const struct ovsrec_port *port_cfg = if_cfg->parent;
1396 struct netdev *netdev;
1397 struct iface *iface;
1402 /* Do the bits that can fail up front.
1404 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1405 * internal datastructures may not be consistent. Eventually, when port
1406 * additions and deletions are cheaper, these calls should be removed. */
1408 ovs_assert(!iface_lookup(br, iface_cfg->name));
1409 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1412 iface_set_ofport(iface_cfg, -1);
1413 iface_clear_db_record(iface_cfg);
1418 /* Get or create the port structure. */
1419 port = port_lookup(br, port_cfg->name);
1421 port = port_create(br, port_cfg);
1424 /* Create the iface structure. */
1425 iface = xzalloc(sizeof *iface);
1426 list_push_back(&port->ifaces, &iface->port_elem);
1427 hmap_insert(&br->iface_by_name, &iface->name_node,
1428 hash_string(iface_cfg->name, 0));
1430 iface->name = xstrdup(iface_cfg->name);
1431 iface->ofp_port = -1;
1432 iface->netdev = netdev;
1433 iface->type = iface_get_type(iface_cfg, br->cfg);
1434 iface->cfg = iface_cfg;
1436 iface_set_ofp_port(iface, ofp_port);
1438 /* Populate initial status in database. */
1439 iface_refresh_stats(iface);
1440 iface_refresh_status(iface);
1442 /* Add bond fake iface if necessary. */
1443 if (port_is_bond_fake_iface(port)) {
1444 struct ofproto_port ofproto_port;
1446 if (ofproto_port_query_by_name(br->ofproto, port->name,
1448 struct netdev *netdev;
1451 error = netdev_open(port->name, "internal", &netdev);
1453 uint16_t fake_ofp_port = if_cfg->ofport;
1455 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1456 netdev_close(netdev);
1458 VLOG_WARN("could not open network device %s (%s)",
1459 port->name, strerror(error));
1462 /* Already exists, nothing to do. */
1463 ofproto_port_destroy(&ofproto_port);
1468 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1474 /* Set Flow eviction threshold */
1476 bridge_configure_flow_eviction_threshold(struct bridge *br)
1478 const char *threshold_str;
1481 threshold_str = smap_get(&br->cfg->other_config,
1482 "flow-eviction-threshold");
1483 if (threshold_str) {
1484 threshold = strtoul(threshold_str, NULL, 10);
1486 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1488 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1491 /* Set forward BPDU option. */
1493 bridge_configure_forward_bpdu(struct bridge *br)
1495 ofproto_set_forward_bpdu(br->ofproto,
1496 smap_get_bool(&br->cfg->other_config,
1501 /* Set MAC learning table configuration for 'br'. */
1503 bridge_configure_mac_table(struct bridge *br)
1505 const char *idle_time_str;
1508 const char *mac_table_size_str;
1511 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1512 idle_time = (idle_time_str && atoi(idle_time_str)
1513 ? atoi(idle_time_str)
1514 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1516 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1517 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1518 ? atoi(mac_table_size_str)
1521 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1525 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1526 struct iface **hw_addr_iface)
1528 struct hmapx mirror_output_ports;
1531 bool found_addr = false;
1535 *hw_addr_iface = NULL;
1537 /* Did the user request a particular MAC? */
1538 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1539 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1540 if (eth_addr_is_multicast(ea)) {
1541 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1542 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1543 } else if (eth_addr_is_zero(ea)) {
1544 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1550 /* Mirror output ports don't participate in picking the local hardware
1551 * address. ofproto can't help us find out whether a given port is a
1552 * mirror output because we haven't configured mirrors yet, so we need to
1553 * accumulate them ourselves. */
1554 hmapx_init(&mirror_output_ports);
1555 for (i = 0; i < br->cfg->n_mirrors; i++) {
1556 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1557 if (m->output_port) {
1558 hmapx_add(&mirror_output_ports, m->output_port);
1562 /* Otherwise choose the minimum non-local MAC address among all of the
1564 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1565 uint8_t iface_ea[ETH_ADDR_LEN];
1566 struct iface *candidate;
1567 struct iface *iface;
1569 /* Mirror output ports don't participate. */
1570 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1574 /* Choose the MAC address to represent the port. */
1576 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1577 /* Find the interface with this Ethernet address (if any) so that
1578 * we can provide the correct devname to the caller. */
1579 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1580 uint8_t candidate_ea[ETH_ADDR_LEN];
1581 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1582 && eth_addr_equals(iface_ea, candidate_ea)) {
1587 /* Choose the interface whose MAC address will represent the port.
1588 * The Linux kernel bonding code always chooses the MAC address of
1589 * the first slave added to a bond, and the Fedora networking
1590 * scripts always add slaves to a bond in alphabetical order, so
1591 * for compatibility we choose the interface with the name that is
1592 * first in alphabetical order. */
1593 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1594 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1599 /* The local port doesn't count (since we're trying to choose its
1600 * MAC address anyway). */
1601 if (iface->ofp_port == OFPP_LOCAL) {
1606 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1612 /* Compare against our current choice. */
1613 if (!eth_addr_is_multicast(iface_ea) &&
1614 !eth_addr_is_local(iface_ea) &&
1615 !eth_addr_is_reserved(iface_ea) &&
1616 !eth_addr_is_zero(iface_ea) &&
1617 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1619 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1620 *hw_addr_iface = iface;
1626 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1627 *hw_addr_iface = NULL;
1630 hmapx_destroy(&mirror_output_ports);
1633 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1634 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1635 * an interface on 'br', then that interface must be passed in as
1636 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1637 * 'hw_addr_iface' must be passed in as a null pointer. */
1639 bridge_pick_datapath_id(struct bridge *br,
1640 const uint8_t bridge_ea[ETH_ADDR_LEN],
1641 struct iface *hw_addr_iface)
1644 * The procedure for choosing a bridge MAC address will, in the most
1645 * ordinary case, also choose a unique MAC that we can use as a datapath
1646 * ID. In some special cases, though, multiple bridges will end up with
1647 * the same MAC address. This is OK for the bridges, but it will confuse
1648 * the OpenFlow controller, because each datapath needs a unique datapath
1651 * Datapath IDs must be unique. It is also very desirable that they be
1652 * stable from one run to the next, so that policy set on a datapath
1655 const char *datapath_id;
1658 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1659 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1663 if (!hw_addr_iface) {
1665 * A purely internal bridge, that is, one that has no non-virtual
1666 * network devices on it at all, is difficult because it has no
1667 * natural unique identifier at all.
1669 * When the host is a XenServer, we handle this case by hashing the
1670 * host's UUID with the name of the bridge. Names of bridges are
1671 * persistent across XenServer reboots, although they can be reused if
1672 * an internal network is destroyed and then a new one is later
1673 * created, so this is fairly effective.
1675 * When the host is not a XenServer, we punt by using a random MAC
1676 * address on each run.
1678 const char *host_uuid = xenserver_get_host_uuid();
1680 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1681 dpid = dpid_from_hash(combined, strlen(combined));
1687 return eth_addr_to_uint64(bridge_ea);
1691 dpid_from_hash(const void *data, size_t n)
1693 uint8_t hash[SHA1_DIGEST_SIZE];
1695 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1696 sha1_bytes(data, n, hash);
1697 eth_addr_mark_random(hash);
1698 return eth_addr_to_uint64(hash);
1702 iface_refresh_status(struct iface *iface)
1706 enum netdev_features current;
1710 uint8_t mac[ETH_ADDR_LEN];
1713 if (iface_is_synthetic(iface)) {
1719 if (!netdev_get_status(iface->netdev, &smap)) {
1720 ovsrec_interface_set_status(iface->cfg, &smap);
1722 ovsrec_interface_set_status(iface->cfg, NULL);
1725 smap_destroy(&smap);
1727 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1728 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1730 ovsrec_interface_set_duplex(iface->cfg,
1731 netdev_features_is_full_duplex(current)
1733 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1735 ovsrec_interface_set_duplex(iface->cfg, NULL);
1736 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1739 error = netdev_get_mtu(iface->netdev, &mtu);
1742 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1744 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1747 error = netdev_get_etheraddr(iface->netdev, mac);
1749 char mac_string[32];
1751 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1752 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1754 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1758 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1761 iface_refresh_cfm_stats(struct iface *iface)
1763 const struct ovsrec_interface *cfg = iface->cfg;
1764 struct ofproto_cfm_status status;
1766 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1767 iface->ofp_port, &status)) {
1768 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1769 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1770 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1771 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1772 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1774 const char *reasons[CFM_FAULT_N_REASONS];
1775 int64_t cfm_health = status.health;
1776 bool faulted = status.faults != 0;
1779 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1782 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1783 int reason = 1 << i;
1784 if (status.faults & reason) {
1785 reasons[j++] = cfm_fault_reason_to_str(reason);
1788 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1790 if (status.remote_opstate >= 0) {
1791 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1792 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1794 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1797 ovsrec_interface_set_cfm_remote_mpids(cfg,
1798 (const int64_t *)status.rmps,
1800 if (cfm_health >= 0) {
1801 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1803 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1809 iface_refresh_stats(struct iface *iface)
1811 #define IFACE_STATS \
1812 IFACE_STAT(rx_packets, "rx_packets") \
1813 IFACE_STAT(tx_packets, "tx_packets") \
1814 IFACE_STAT(rx_bytes, "rx_bytes") \
1815 IFACE_STAT(tx_bytes, "tx_bytes") \
1816 IFACE_STAT(rx_dropped, "rx_dropped") \
1817 IFACE_STAT(tx_dropped, "tx_dropped") \
1818 IFACE_STAT(rx_errors, "rx_errors") \
1819 IFACE_STAT(tx_errors, "tx_errors") \
1820 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1821 IFACE_STAT(rx_over_errors, "rx_over_err") \
1822 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1823 IFACE_STAT(collisions, "collisions")
1825 #define IFACE_STAT(MEMBER, NAME) NAME,
1826 static char *keys[] = { IFACE_STATS };
1828 int64_t values[ARRAY_SIZE(keys)];
1831 struct netdev_stats stats;
1833 if (iface_is_synthetic(iface)) {
1837 /* Intentionally ignore return value, since errors will set 'stats' to
1838 * all-1s, and we will deal with that correctly below. */
1839 netdev_get_stats(iface->netdev, &stats);
1841 /* Copy statistics into values[] array. */
1843 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1846 ovs_assert(i == ARRAY_SIZE(keys));
1848 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1854 br_refresh_stp_status(struct bridge *br)
1856 struct smap smap = SMAP_INITIALIZER(&smap);
1857 struct ofproto *ofproto = br->ofproto;
1858 struct ofproto_stp_status status;
1860 if (ofproto_get_stp_status(ofproto, &status)) {
1864 if (!status.enabled) {
1865 ovsrec_bridge_set_status(br->cfg, NULL);
1869 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1870 STP_ID_ARGS(status.bridge_id));
1871 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1872 STP_ID_ARGS(status.designated_root));
1873 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1875 ovsrec_bridge_set_status(br->cfg, &smap);
1876 smap_destroy(&smap);
1880 port_refresh_stp_status(struct port *port)
1882 struct ofproto *ofproto = port->bridge->ofproto;
1883 struct iface *iface;
1884 struct ofproto_port_stp_status status;
1886 int64_t int_values[3];
1889 if (port_is_synthetic(port)) {
1893 /* STP doesn't currently support bonds. */
1894 if (!list_is_singleton(&port->ifaces)) {
1895 ovsrec_port_set_status(port->cfg, NULL);
1899 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1901 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1905 if (!status.enabled) {
1906 ovsrec_port_set_status(port->cfg, NULL);
1907 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1911 /* Set Status column. */
1913 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1914 smap_add(&smap, "stp_state", stp_state_name(status.state));
1915 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1916 smap_add(&smap, "stp_role", stp_role_name(status.role));
1917 ovsrec_port_set_status(port->cfg, &smap);
1918 smap_destroy(&smap);
1920 /* Set Statistics column. */
1921 keys[0] = "stp_tx_count";
1922 int_values[0] = status.tx_count;
1923 keys[1] = "stp_rx_count";
1924 int_values[1] = status.rx_count;
1925 keys[2] = "stp_error_count";
1926 int_values[2] = status.error_count;
1928 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1929 ARRAY_SIZE(int_values));
1933 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1935 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1939 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1941 bool enable = enable_system_stats(cfg);
1943 system_stats_enable(enable);
1945 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1950 run_system_stats(void)
1952 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1955 stats = system_stats_run();
1957 struct ovsdb_idl_txn *txn;
1958 struct ovsdb_datum datum;
1960 txn = ovsdb_idl_txn_create(idl);
1961 ovsdb_datum_from_smap(&datum, stats);
1962 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1964 ovsdb_idl_txn_commit(txn);
1965 ovsdb_idl_txn_destroy(txn);
1971 static inline const char *
1972 ofp12_controller_role_to_str(enum ofp12_controller_role role)
1975 case OFPCR12_ROLE_EQUAL:
1977 case OFPCR12_ROLE_MASTER:
1979 case OFPCR12_ROLE_SLAVE:
1981 case OFPCR12_ROLE_NOCHANGE:
1983 return "*** INVALID ROLE ***";
1988 refresh_controller_status(void)
1992 const struct ovsrec_controller *cfg;
1996 /* Accumulate status for controllers on all bridges. */
1997 HMAP_FOR_EACH (br, node, &all_bridges) {
1998 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2001 /* Update each controller in the database with current status. */
2002 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2003 struct ofproto_controller_info *cinfo =
2004 shash_find_data(&info, cfg->target);
2007 struct smap smap = SMAP_INITIALIZER(&smap);
2008 const char **values = cinfo->pairs.values;
2009 const char **keys = cinfo->pairs.keys;
2012 for (i = 0; i < cinfo->pairs.n; i++) {
2013 smap_add(&smap, keys[i], values[i]);
2016 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2017 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2019 ovsrec_controller_set_status(cfg, &smap);
2020 smap_destroy(&smap);
2022 ovsrec_controller_set_is_connected(cfg, false);
2023 ovsrec_controller_set_role(cfg, NULL);
2024 ovsrec_controller_set_status(cfg, NULL);
2028 ofproto_free_ofproto_controller_info(&info);
2033 * Some information in the database must be kept as up-to-date as possible to
2034 * allow controllers to respond rapidly to network outages. We call these
2035 * statistics "instant" stats.
2037 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2038 * assuming that they've changed. The only means we have to determine whether
2039 * they have changed are:
2041 * - Try to commit changes to the database. If nothing changed, then
2042 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2045 * - instant_stats_run() is called late in the run loop, after anything that
2046 * might change any of the instant stats.
2048 * We use these two facts together to avoid waking the process up every
2049 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2052 /* Minimum interval between writing updates to the instant stats to the
2054 #define INSTANT_INTERVAL_MSEC 100
2056 /* Current instant stats database transaction, NULL if there is no ongoing
2058 static struct ovsdb_idl_txn *instant_txn;
2060 /* Next time (in msec on monotonic clock) at which we will update the instant
2062 static long long int instant_next_txn = LLONG_MIN;
2064 /* True if the run loop has run since we last saw that the instant stats were
2065 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2066 * to refresh the instant stats. */
2067 static bool instant_stats_could_have_changed;
2070 instant_stats_run(void)
2072 enum ovsdb_idl_txn_status status;
2074 instant_stats_could_have_changed = true;
2079 if (time_msec() < instant_next_txn) {
2082 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2084 instant_txn = ovsdb_idl_txn_create(idl);
2085 HMAP_FOR_EACH (br, node, &all_bridges) {
2086 struct iface *iface;
2089 br_refresh_stp_status(br);
2091 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2092 port_refresh_stp_status(port);
2095 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2096 enum netdev_flags flags;
2097 const char *link_state;
2098 int64_t link_resets;
2101 if (iface_is_synthetic(iface)) {
2105 current = ofproto_port_is_lacp_current(br->ofproto,
2109 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2111 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2114 error = netdev_get_flags(iface->netdev, &flags);
2116 const char *state = flags & NETDEV_UP ? "up" : "down";
2117 ovsrec_interface_set_admin_state(iface->cfg, state);
2119 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2122 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2123 ovsrec_interface_set_link_state(iface->cfg, link_state);
2125 link_resets = netdev_get_carrier_resets(iface->netdev);
2126 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2128 iface_refresh_cfm_stats(iface);
2133 status = ovsdb_idl_txn_commit(instant_txn);
2134 if (status != TXN_INCOMPLETE) {
2135 ovsdb_idl_txn_destroy(instant_txn);
2138 if (status == TXN_UNCHANGED) {
2139 instant_stats_could_have_changed = false;
2144 instant_stats_wait(void)
2147 ovsdb_idl_txn_wait(instant_txn);
2148 } else if (instant_stats_could_have_changed) {
2149 poll_timer_wait_until(instant_next_txn);
2153 /* Performs periodic activity required by bridges that needs to be done with
2154 * the least possible latency.
2156 * It makes sense to call this function a couple of times per poll loop, to
2157 * provide a significant performance boost on some benchmarks with ofprotos
2158 * that use the ofproto-dpif implementation. */
2160 bridge_run_fast(void)
2167 ofproto_enumerate_types(&types);
2168 SSET_FOR_EACH (type, &types) {
2169 ofproto_type_run_fast(type);
2171 sset_destroy(&types);
2173 HMAP_FOR_EACH (br, node, &all_bridges) {
2174 ofproto_run_fast(br->ofproto);
2181 static struct ovsrec_open_vswitch null_cfg;
2182 const struct ovsrec_open_vswitch *cfg;
2183 struct ovsdb_idl_txn *reconf_txn = NULL;
2187 bool vlan_splinters_changed;
2190 ovsrec_open_vswitch_init(&null_cfg);
2192 /* (Re)configure if necessary. */
2193 if (!reconfiguring) {
2196 if (ovsdb_idl_is_lock_contended(idl)) {
2197 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2198 struct bridge *br, *next_br;
2200 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2201 "disabling this process until it goes away");
2203 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2207 } else if (!ovsdb_idl_has_lock(idl)) {
2211 cfg = ovsrec_open_vswitch_first(idl);
2213 /* Initialize the ofproto library. This only needs to run once, but
2214 * it must be done after the configuration is set. If the
2215 * initialization has already occurred, bridge_init_ofproto()
2216 * returns immediately. */
2217 bridge_init_ofproto(cfg);
2219 /* Let each datapath type do the work that it needs to do. */
2221 ofproto_enumerate_types(&types);
2222 SSET_FOR_EACH (type, &types) {
2223 ofproto_type_run(type);
2225 sset_destroy(&types);
2227 /* Let each bridge do the work that it needs to do. */
2228 HMAP_FOR_EACH (br, node, &all_bridges) {
2229 ofproto_run(br->ofproto);
2232 /* Re-configure SSL. We do this on every trip through the main loop,
2233 * instead of just when the database changes, because the contents of the
2234 * key and certificate files can change without the database changing.
2236 * We do this before bridge_reconfigure() because that function might
2237 * initiate SSL connections and thus requires SSL to be configured. */
2238 if (cfg && cfg->ssl) {
2239 const struct ovsrec_ssl *ssl = cfg->ssl;
2241 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2242 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2245 if (!reconfiguring) {
2246 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2247 * usage has changed. */
2248 vlan_splinters_changed = false;
2249 if (vlan_splinters_enabled_anywhere) {
2250 HMAP_FOR_EACH (br, node, &all_bridges) {
2251 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2252 vlan_splinters_changed = true;
2258 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2259 idl_seqno = ovsdb_idl_get_seqno(idl);
2261 reconf_txn = ovsdb_idl_txn_create(idl);
2262 bridge_reconfigure(cfg);
2264 /* We still need to reconfigure to avoid dangling pointers to
2265 * now-destroyed ovsrec structures inside bridge data. */
2266 bridge_reconfigure(&null_cfg);
2271 if (reconfiguring) {
2273 reconf_txn = ovsdb_idl_txn_create(idl);
2276 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2277 reconfiguring = false;
2280 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2283 /* If we are completing our initial configuration for this run
2284 * of ovs-vswitchd, then keep the transaction around to monitor
2285 * it for completion. */
2286 if (!initial_config_done) {
2287 initial_config_done = true;
2288 daemonize_txn = reconf_txn;
2295 ovsdb_idl_txn_commit(reconf_txn);
2296 ovsdb_idl_txn_destroy(reconf_txn);
2300 if (daemonize_txn) {
2301 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2302 if (status != TXN_INCOMPLETE) {
2303 ovsdb_idl_txn_destroy(daemonize_txn);
2304 daemonize_txn = NULL;
2306 /* ovs-vswitchd has completed initialization, so allow the
2307 * process that forked us to exit successfully. */
2308 daemonize_complete();
2310 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2314 /* Refresh interface and mirror stats if necessary. */
2315 if (time_msec() >= iface_stats_timer) {
2317 struct ovsdb_idl_txn *txn;
2319 txn = ovsdb_idl_txn_create(idl);
2320 HMAP_FOR_EACH (br, node, &all_bridges) {
2324 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2325 struct iface *iface;
2327 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2328 iface_refresh_stats(iface);
2329 iface_refresh_status(iface);
2333 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2334 mirror_refresh_stats(m);
2338 refresh_controller_status();
2339 ovsdb_idl_txn_commit(txn);
2340 ovsdb_idl_txn_destroy(txn); /* XXX */
2343 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2347 instant_stats_run();
2356 ovsdb_idl_wait(idl);
2357 if (daemonize_txn) {
2358 ovsdb_idl_txn_wait(daemonize_txn);
2361 if (reconfiguring) {
2362 poll_immediate_wake();
2366 ofproto_enumerate_types(&types);
2367 SSET_FOR_EACH (type, &types) {
2368 ofproto_type_wait(type);
2370 sset_destroy(&types);
2372 if (!hmap_is_empty(&all_bridges)) {
2375 HMAP_FOR_EACH (br, node, &all_bridges) {
2376 ofproto_wait(br->ofproto);
2378 poll_timer_wait_until(iface_stats_timer);
2381 system_stats_wait();
2382 instant_stats_wait();
2385 /* Adds some memory usage statistics for bridges into 'usage', for use with
2386 * memory_report(). */
2388 bridge_get_memory_usage(struct simap *usage)
2392 HMAP_FOR_EACH (br, node, &all_bridges) {
2393 ofproto_get_memory_usage(br->ofproto, usage);
2397 /* QoS unixctl user interface functions. */
2399 struct qos_unixctl_show_cbdata {
2401 struct iface *iface;
2405 qos_unixctl_show_cb(unsigned int queue_id,
2406 const struct smap *details,
2409 struct qos_unixctl_show_cbdata *data = aux;
2410 struct ds *ds = data->ds;
2411 struct iface *iface = data->iface;
2412 struct netdev_queue_stats stats;
2413 struct smap_node *node;
2416 ds_put_cstr(ds, "\n");
2418 ds_put_format(ds, "Queue %u:\n", queue_id);
2420 ds_put_cstr(ds, "Default:\n");
2423 SMAP_FOR_EACH (node, details) {
2424 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2427 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2429 if (stats.tx_packets != UINT64_MAX) {
2430 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2433 if (stats.tx_bytes != UINT64_MAX) {
2434 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2437 if (stats.tx_errors != UINT64_MAX) {
2438 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2441 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2442 queue_id, strerror(error));
2447 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2448 const char *argv[], void *aux OVS_UNUSED)
2450 struct ds ds = DS_EMPTY_INITIALIZER;
2451 struct smap smap = SMAP_INITIALIZER(&smap);
2452 struct iface *iface;
2454 struct smap_node *node;
2455 struct qos_unixctl_show_cbdata data;
2458 iface = iface_find(argv[1]);
2460 unixctl_command_reply_error(conn, "no such interface");
2464 netdev_get_qos(iface->netdev, &type, &smap);
2466 if (*type != '\0') {
2467 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2469 SMAP_FOR_EACH (node, &smap) {
2470 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2475 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2478 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2480 unixctl_command_reply(conn, ds_cstr(&ds));
2482 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2483 unixctl_command_reply_error(conn, ds_cstr(&ds));
2486 smap_destroy(&smap);
2490 /* Bridge reconfiguration functions. */
2492 bridge_create(const struct ovsrec_bridge *br_cfg)
2496 ovs_assert(!bridge_lookup(br_cfg->name));
2497 br = xzalloc(sizeof *br);
2499 br->name = xstrdup(br_cfg->name);
2500 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2503 /* Derive the default Ethernet address from the bridge's UUID. This should
2504 * be unique and it will be stable between ovs-vswitchd runs. */
2505 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2506 eth_addr_mark_random(br->default_ea);
2508 hmap_init(&br->ports);
2509 hmap_init(&br->ifaces);
2510 hmap_init(&br->iface_by_name);
2511 hmap_init(&br->mirrors);
2513 hmap_init(&br->if_cfg_todo);
2514 list_init(&br->ofpp_garbage);
2516 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2520 bridge_destroy(struct bridge *br)
2523 struct mirror *mirror, *next_mirror;
2524 struct port *port, *next_port;
2525 struct if_cfg *if_cfg, *next_if_cfg;
2526 struct ofpp_garbage *garbage, *next_garbage;
2528 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2531 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2532 mirror_destroy(mirror);
2534 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2535 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2538 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2539 &br->ofpp_garbage) {
2540 list_remove(&garbage->list_node);
2544 hmap_remove(&all_bridges, &br->node);
2545 ofproto_destroy(br->ofproto);
2546 hmap_destroy(&br->ifaces);
2547 hmap_destroy(&br->ports);
2548 hmap_destroy(&br->iface_by_name);
2549 hmap_destroy(&br->mirrors);
2550 hmap_destroy(&br->if_cfg_todo);
2557 static struct bridge *
2558 bridge_lookup(const char *name)
2562 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2563 if (!strcmp(br->name, name)) {
2570 /* Handle requests for a listing of all flows known by the OpenFlow
2571 * stack, including those normally hidden. */
2573 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2574 const char *argv[], void *aux OVS_UNUSED)
2579 br = bridge_lookup(argv[1]);
2581 unixctl_command_reply_error(conn, "Unknown bridge");
2586 ofproto_get_all_flows(br->ofproto, &results);
2588 unixctl_command_reply(conn, ds_cstr(&results));
2589 ds_destroy(&results);
2592 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2593 * connections and reconnect. If BRIDGE is not specified, then all bridges
2594 * drop their controller connections and reconnect. */
2596 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2597 const char *argv[], void *aux OVS_UNUSED)
2601 br = bridge_lookup(argv[1]);
2603 unixctl_command_reply_error(conn, "Unknown bridge");
2606 ofproto_reconnect_controllers(br->ofproto);
2608 HMAP_FOR_EACH (br, node, &all_bridges) {
2609 ofproto_reconnect_controllers(br->ofproto);
2612 unixctl_command_reply(conn, NULL);
2616 bridge_get_controllers(const struct bridge *br,
2617 struct ovsrec_controller ***controllersp)
2619 struct ovsrec_controller **controllers;
2620 size_t n_controllers;
2622 controllers = br->cfg->controller;
2623 n_controllers = br->cfg->n_controller;
2625 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2631 *controllersp = controllers;
2633 return n_controllers;
2637 bridge_queue_if_cfg(struct bridge *br,
2638 const struct ovsrec_interface *cfg,
2639 const struct ovsrec_port *parent)
2641 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2644 if_cfg->parent = parent;
2645 if_cfg->ofport = iface_pick_ofport(cfg);
2646 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2647 hash_string(if_cfg->cfg->name, 0));
2650 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2651 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2652 * 'br' needs to complete its configuration. */
2654 bridge_add_del_ports(struct bridge *br,
2655 const unsigned long int *splinter_vlans)
2657 struct shash_node *port_node;
2658 struct port *port, *next;
2659 struct shash new_ports;
2662 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2664 /* Collect new ports. */
2665 shash_init(&new_ports);
2666 for (i = 0; i < br->cfg->n_ports; i++) {
2667 const char *name = br->cfg->ports[i]->name;
2668 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2669 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2673 if (bridge_get_controllers(br, NULL)
2674 && !shash_find(&new_ports, br->name)) {
2675 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2676 br->name, br->name);
2678 ovsrec_interface_init(&br->synth_local_iface);
2679 ovsrec_port_init(&br->synth_local_port);
2681 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2682 br->synth_local_port.n_interfaces = 1;
2683 br->synth_local_port.name = br->name;
2685 br->synth_local_iface.name = br->name;
2686 br->synth_local_iface.type = "internal";
2688 br->synth_local_ifacep = &br->synth_local_iface;
2690 shash_add(&new_ports, br->name, &br->synth_local_port);
2693 if (splinter_vlans) {
2694 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2697 /* Get rid of deleted ports.
2698 * Get rid of deleted interfaces on ports that still exist. */
2699 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2700 port->cfg = shash_find_data(&new_ports, port->name);
2704 port_del_ifaces(port);
2708 /* Update iface->cfg and iface->type in interfaces that still exist.
2709 * Add new interfaces to creation queue. */
2710 SHASH_FOR_EACH (port_node, &new_ports) {
2711 const struct ovsrec_port *port = port_node->data;
2714 for (i = 0; i < port->n_interfaces; i++) {
2715 const struct ovsrec_interface *cfg = port->interfaces[i];
2716 struct iface *iface = iface_lookup(br, cfg->name);
2717 const char *type = iface_get_type(cfg, br->cfg);
2722 } else if (!strcmp(type, "null")) {
2723 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2724 " may be removed in February 2013. Please email"
2725 " dev@openvswitch.org with concerns.",
2728 bridge_queue_if_cfg(br, cfg, port);
2733 shash_destroy(&new_ports);
2736 /* Initializes 'oc' appropriately as a management service controller for
2739 * The caller must free oc->target when it is no longer needed. */
2741 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2742 struct ofproto_controller *oc)
2744 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2745 oc->max_backoff = 0;
2746 oc->probe_interval = 60;
2747 oc->band = OFPROTO_OUT_OF_BAND;
2749 oc->burst_limit = 0;
2750 oc->enable_async_msgs = true;
2753 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2755 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2756 struct ofproto_controller *oc)
2760 oc->target = c->target;
2761 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2762 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2763 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2764 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2765 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2766 oc->burst_limit = (c->controller_burst_limit
2767 ? *c->controller_burst_limit : 0);
2768 oc->enable_async_msgs = (!c->enable_async_messages
2769 || *c->enable_async_messages);
2770 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2771 if (dscp < 0 || dscp > 63) {
2772 dscp = DSCP_DEFAULT;
2777 /* Configures the IP stack for 'br''s local interface properly according to the
2778 * configuration in 'c'. */
2780 bridge_configure_local_iface_netdev(struct bridge *br,
2781 struct ovsrec_controller *c)
2783 struct netdev *netdev;
2784 struct in_addr mask, gateway;
2786 struct iface *local_iface;
2789 /* If there's no local interface or no IP address, give up. */
2790 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2791 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2795 /* Bring up the local interface. */
2796 netdev = local_iface->netdev;
2797 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2799 /* Configure the IP address and netmask. */
2800 if (!c->local_netmask
2801 || !inet_aton(c->local_netmask, &mask)
2803 mask.s_addr = guess_netmask(ip.s_addr);
2805 if (!netdev_set_in4(netdev, ip, mask)) {
2806 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2807 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2810 /* Configure the default gateway. */
2811 if (c->local_gateway
2812 && inet_aton(c->local_gateway, &gateway)
2813 && gateway.s_addr) {
2814 if (!netdev_add_router(netdev, gateway)) {
2815 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2816 br->name, IP_ARGS(gateway.s_addr));
2821 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2822 * in either string are treated as equal to any number of slashes in the other,
2823 * e.g. "x///y" is equal to "x/y".
2825 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2826 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2827 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2828 * 'b' against a prefix of 'a'.
2831 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2833 const char *b_start = b;
2835 if (b - b_start >= b_stoplen) {
2837 } else if (*a != *b) {
2839 } else if (*a == '/') {
2840 a += strspn(a, "/");
2841 b += strspn(b, "/");
2842 } else if (*a == '\0') {
2852 bridge_configure_remotes(struct bridge *br,
2853 const struct sockaddr_in *managers, size_t n_managers)
2855 bool disable_in_band;
2857 struct ovsrec_controller **controllers;
2858 size_t n_controllers;
2860 enum ofproto_fail_mode fail_mode;
2862 struct ofproto_controller *ocs;
2866 /* Check if we should disable in-band control on this bridge. */
2867 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2870 /* Set OpenFlow queue ID for in-band control. */
2871 ofproto_set_in_band_queue(br->ofproto,
2872 smap_get_int(&br->cfg->other_config,
2873 "in-band-queue", -1));
2875 if (disable_in_band) {
2876 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2878 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2881 n_controllers = bridge_get_controllers(br, &controllers);
2883 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2886 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2887 for (i = 0; i < n_controllers; i++) {
2888 struct ovsrec_controller *c = controllers[i];
2890 if (!strncmp(c->target, "punix:", 6)
2891 || !strncmp(c->target, "unix:", 5)) {
2892 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2895 if (!strncmp(c->target, "unix:", 5)) {
2896 /* Connect to a listening socket */
2897 whitelist = xasprintf("unix:%s/", ovs_rundir());
2898 if (strchr(c->target, '/') &&
2899 !equal_pathnames(c->target, whitelist,
2900 strlen(whitelist))) {
2901 /* Absolute path specified, but not in ovs_rundir */
2902 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2903 "controller \"%s\" due to possibility for "
2904 "remote exploit. Instead, specify socket "
2905 "in whitelisted \"%s\" or connect to "
2906 "\"unix:%s/%s.mgmt\" (which is always "
2907 "available without special configuration).",
2908 br->name, c->target, whitelist,
2909 ovs_rundir(), br->name);
2914 whitelist = xasprintf("punix:%s/%s.controller",
2915 ovs_rundir(), br->name);
2916 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2917 /* Prevent remote ovsdb-server users from accessing
2918 * arbitrary Unix domain sockets and overwriting arbitrary
2920 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2921 "controller \"%s\" due to possibility of "
2922 "overwriting local files. Instead, specify "
2923 "whitelisted \"%s\" or connect to "
2924 "\"unix:%s/%s.mgmt\" (which is always "
2925 "available without special configuration).",
2926 br->name, c->target, whitelist,
2927 ovs_rundir(), br->name);
2936 bridge_configure_local_iface_netdev(br, c);
2937 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2938 if (disable_in_band) {
2939 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2944 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2945 bridge_get_allowed_versions(br));
2946 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2949 /* Set the fail-mode. */
2950 fail_mode = !br->cfg->fail_mode
2951 || !strcmp(br->cfg->fail_mode, "standalone")
2952 ? OFPROTO_FAIL_STANDALONE
2953 : OFPROTO_FAIL_SECURE;
2954 ofproto_set_fail_mode(br->ofproto, fail_mode);
2956 /* Configure OpenFlow controller connection snooping. */
2957 if (!ofproto_has_snoops(br->ofproto)) {
2961 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2962 ovs_rundir(), br->name));
2963 ofproto_set_snoops(br->ofproto, &snoops);
2964 sset_destroy(&snoops);
2969 bridge_configure_tables(struct bridge *br)
2971 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2975 n_tables = ofproto_get_n_tables(br->ofproto);
2977 for (i = 0; i < n_tables; i++) {
2978 struct ofproto_table_settings s;
2981 s.max_flows = UINT_MAX;
2985 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
2986 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
2989 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
2990 s.max_flows = *cfg->flow_limit;
2992 if (cfg->overflow_policy
2993 && !strcmp(cfg->overflow_policy, "evict")) {
2996 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
2997 for (k = 0; k < cfg->n_groups; k++) {
2998 const char *string = cfg->groups[k];
3001 msg = mf_parse_subfield__(&s.groups[k], &string);
3003 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3004 "'groups' (%s)", br->name, i, msg);
3006 } else if (*string) {
3007 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3008 "element '%s' contains trailing garbage",
3009 br->name, i, cfg->groups[k]);
3017 ofproto_configure_table(br->ofproto, i, &s);
3021 for (; j < br->cfg->n_flow_tables; j++) {
3022 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3023 "%"PRId64" not supported by this datapath", br->name,
3024 br->cfg->key_flow_tables[j]);
3029 bridge_configure_dp_desc(struct bridge *br)
3031 ofproto_set_dp_desc(br->ofproto,
3032 smap_get(&br->cfg->other_config, "dp-desc"));
3035 /* Port functions. */
3037 static struct port *
3038 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3042 port = xzalloc(sizeof *port);
3044 port->name = xstrdup(cfg->name);
3046 list_init(&port->ifaces);
3048 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3052 /* Deletes interfaces from 'port' that are no longer configured for it. */
3054 port_del_ifaces(struct port *port)
3056 struct iface *iface, *next;
3057 struct sset new_ifaces;
3060 /* Collect list of new interfaces. */
3061 sset_init(&new_ifaces);
3062 for (i = 0; i < port->cfg->n_interfaces; i++) {
3063 const char *name = port->cfg->interfaces[i]->name;
3064 const char *type = port->cfg->interfaces[i]->type;
3065 if (strcmp(type, "null")) {
3066 sset_add(&new_ifaces, name);
3070 /* Get rid of deleted interfaces. */
3071 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3072 if (!sset_contains(&new_ifaces, iface->name)) {
3073 iface_destroy(iface);
3077 sset_destroy(&new_ifaces);
3081 port_destroy(struct port *port)
3084 struct bridge *br = port->bridge;
3085 struct iface *iface, *next;
3088 ofproto_bundle_unregister(br->ofproto, port);
3091 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3092 iface_destroy(iface);
3095 hmap_remove(&br->ports, &port->hmap_node);
3101 static struct port *
3102 port_lookup(const struct bridge *br, const char *name)
3106 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3108 if (!strcmp(port->name, name)) {
3116 enable_lacp(struct port *port, bool *activep)
3118 if (!port->cfg->lacp) {
3119 /* XXX when LACP implementation has been sufficiently tested, enable by
3120 * default and make active on bonded ports. */
3122 } else if (!strcmp(port->cfg->lacp, "off")) {
3124 } else if (!strcmp(port->cfg->lacp, "active")) {
3127 } else if (!strcmp(port->cfg->lacp, "passive")) {
3131 VLOG_WARN("port %s: unknown LACP mode %s",
3132 port->name, port->cfg->lacp);
3137 static struct lacp_settings *
3138 port_configure_lacp(struct port *port, struct lacp_settings *s)
3140 const char *lacp_time, *system_id;
3143 if (!enable_lacp(port, &s->active)) {
3147 s->name = port->name;
3149 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3151 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3152 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3153 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3154 " address.", port->name, system_id);
3158 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3161 if (eth_addr_is_zero(s->id)) {
3162 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3166 /* Prefer bondable links if unspecified. */
3167 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3169 s->priority = (priority > 0 && priority <= UINT16_MAX
3171 : UINT16_MAX - !list_is_short(&port->ifaces));
3173 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3174 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3179 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3181 int priority, portid, key;
3183 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3184 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3186 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3188 if (portid <= 0 || portid > UINT16_MAX) {
3189 portid = iface->ofp_port;
3192 if (priority <= 0 || priority > UINT16_MAX) {
3193 priority = UINT16_MAX;
3196 if (key < 0 || key > UINT16_MAX) {
3200 s->name = iface->name;
3202 s->priority = priority;
3207 port_configure_bond(struct port *port, struct bond_settings *s)
3209 const char *detect_s;
3210 struct iface *iface;
3211 int miimon_interval;
3213 s->name = port->name;
3215 if (port->cfg->bond_mode) {
3216 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3217 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3218 port->name, port->cfg->bond_mode,
3219 bond_mode_to_string(s->balance));
3222 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3224 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3225 * active-backup. At some point we should remove this warning. */
3226 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3227 " in previous versions, the default bond_mode was"
3228 " balance-slb", port->name,
3229 bond_mode_to_string(s->balance));
3231 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3232 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3233 "please use another bond type or disable flood_vlans",
3237 miimon_interval = smap_get_int(&port->cfg->other_config,
3238 "bond-miimon-interval", 0);
3239 if (miimon_interval <= 0) {
3240 miimon_interval = 200;
3243 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3244 if (!detect_s || !strcmp(detect_s, "carrier")) {
3245 miimon_interval = 0;
3246 } else if (strcmp(detect_s, "miimon")) {
3247 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3248 "defaulting to carrier", port->name, detect_s);
3249 miimon_interval = 0;
3252 s->up_delay = MAX(0, port->cfg->bond_updelay);
3253 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3254 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3255 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3256 "bond-rebalance-interval", 10000);
3257 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3258 s->rebalance_interval = 1000;
3261 s->fake_iface = port->cfg->bond_fake_iface;
3263 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3264 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3268 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3269 * instead of obtaining it from the database. */
3271 port_is_synthetic(const struct port *port)
3273 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3276 /* Interface functions. */
3279 iface_is_internal(const struct ovsrec_interface *iface,
3280 const struct ovsrec_bridge *br)
3282 /* The local port and "internal" ports are always "internal". */
3283 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3286 /* Returns the correct network device type for interface 'iface' in bridge
3289 iface_get_type(const struct ovsrec_interface *iface,
3290 const struct ovsrec_bridge *br)
3294 /* The local port always has type "internal". Other ports take
3295 * their type from the database and default to "system" if none is
3297 if (iface_is_internal(iface, br)) {
3300 type = iface->type[0] ? iface->type : "system";
3303 return ofproto_port_open_type(br->datapath_type, type);
3307 iface_destroy(struct iface *iface)
3310 struct port *port = iface->port;
3311 struct bridge *br = port->bridge;
3313 if (br->ofproto && iface->ofp_port >= 0) {
3314 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3317 if (iface->ofp_port >= 0) {
3318 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3321 list_remove(&iface->port_elem);
3322 hmap_remove(&br->iface_by_name, &iface->name_node);
3324 netdev_close(iface->netdev);
3331 static struct iface *
3332 iface_lookup(const struct bridge *br, const char *name)
3334 struct iface *iface;
3336 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3337 &br->iface_by_name) {
3338 if (!strcmp(iface->name, name)) {
3346 static struct iface *
3347 iface_find(const char *name)
3349 const struct bridge *br;
3351 HMAP_FOR_EACH (br, node, &all_bridges) {
3352 struct iface *iface = iface_lookup(br, name);
3361 static struct if_cfg *
3362 if_cfg_lookup(const struct bridge *br, const char *name)
3364 struct if_cfg *if_cfg;
3366 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3368 if (!strcmp(if_cfg->cfg->name, name)) {
3376 static struct iface *
3377 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3379 struct iface *iface;
3381 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3382 hash_int(ofp_port, 0), &br->ifaces) {
3383 if (iface->ofp_port == ofp_port) {
3390 /* Set Ethernet address of 'iface', if one is specified in the configuration
3393 iface_set_mac(struct iface *iface)
3395 uint8_t ea[ETH_ADDR_LEN];
3397 if (!strcmp(iface->type, "internal")
3398 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3399 if (iface->ofp_port == OFPP_LOCAL) {
3400 VLOG_ERR("interface %s: ignoring mac in Interface record "
3401 "(use Bridge record to set local port's mac)",
3403 } else if (eth_addr_is_multicast(ea)) {
3404 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3407 int error = netdev_set_etheraddr(iface->netdev, ea);
3409 VLOG_ERR("interface %s: setting MAC failed (%s)",
3410 iface->name, strerror(error));
3416 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3418 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3420 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3421 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3425 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3426 * sets the "ofport" field to -1.
3428 * This is appropriate when 'if_cfg''s interface cannot be created or is
3429 * otherwise invalid. */
3431 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3433 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3434 ovsrec_interface_set_status(if_cfg, NULL);
3435 ovsrec_interface_set_admin_state(if_cfg, NULL);
3436 ovsrec_interface_set_duplex(if_cfg, NULL);
3437 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3438 ovsrec_interface_set_link_state(if_cfg, NULL);
3439 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3440 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3441 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3442 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3443 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3444 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3445 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3449 struct iface_delete_queues_cbdata {
3450 struct netdev *netdev;
3451 const struct ovsdb_datum *queues;
3455 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3457 union ovsdb_atom atom;
3459 atom.integer = target;
3460 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3464 iface_delete_queues(unsigned int queue_id,
3465 const struct smap *details OVS_UNUSED, void *cbdata_)
3467 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3469 if (!queue_ids_include(cbdata->queues, queue_id)) {
3470 netdev_delete_queue(cbdata->netdev, queue_id);
3475 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3477 struct ofpbuf queues_buf;
3479 ofpbuf_init(&queues_buf, 0);
3481 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3482 netdev_set_qos(iface->netdev, NULL, NULL);
3484 struct iface_delete_queues_cbdata cbdata;
3488 /* Configure top-level Qos for 'iface'. */
3489 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3491 /* Deconfigure queues that were deleted. */
3492 cbdata.netdev = iface->netdev;
3493 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3495 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3497 /* Configure queues for 'iface'. */
3499 for (i = 0; i < qos->n_queues; i++) {
3500 const struct ovsrec_queue *queue = qos->value_queues[i];
3501 unsigned int queue_id = qos->key_queues[i];
3503 if (queue_id == 0) {
3507 if (queue->n_dscp == 1) {
3508 struct ofproto_port_queue *port_queue;
3510 port_queue = ofpbuf_put_uninit(&queues_buf,
3511 sizeof *port_queue);
3512 port_queue->queue = queue_id;
3513 port_queue->dscp = queue->dscp[0];
3516 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3519 struct smap details;
3521 smap_init(&details);
3522 netdev_set_queue(iface->netdev, 0, &details);
3523 smap_destroy(&details);
3527 if (iface->ofp_port >= 0) {
3528 const struct ofproto_port_queue *port_queues = queues_buf.data;
3529 size_t n_queues = queues_buf.size / sizeof *port_queues;
3531 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3532 port_queues, n_queues);
3535 netdev_set_policing(iface->netdev,
3536 iface->cfg->ingress_policing_rate,
3537 iface->cfg->ingress_policing_burst);
3539 ofpbuf_uninit(&queues_buf);
3543 iface_configure_cfm(struct iface *iface)
3545 const struct ovsrec_interface *cfg = iface->cfg;
3546 const char *opstate_str;
3547 const char *cfm_ccm_vlan;
3548 struct cfm_settings s;
3549 struct smap netdev_args;
3551 if (!cfg->n_cfm_mpid) {
3552 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3556 s.check_tnl_key = false;
3557 smap_init(&netdev_args);
3558 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3559 const char *key = smap_get(&netdev_args, "key");
3560 const char *in_key = smap_get(&netdev_args, "in_key");
3562 s.check_tnl_key = (key && !strcmp(key, "flow"))
3563 || (in_key && !strcmp(in_key, "flow"));
3565 smap_destroy(&netdev_args);
3567 s.mpid = *cfg->cfm_mpid;
3568 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3569 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3570 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3572 if (s.interval <= 0) {
3576 if (!cfm_ccm_vlan) {
3578 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3579 s.ccm_vlan = CFM_RANDOM_VLAN;
3581 s.ccm_vlan = atoi(cfm_ccm_vlan);
3582 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3587 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3590 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3591 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3593 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3596 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3597 * instead of obtaining it from the database. */
3599 iface_is_synthetic(const struct iface *iface)
3601 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3605 iface_pick_ofport(const struct ovsrec_interface *cfg)
3607 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3608 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3612 /* Port mirroring. */
3614 static struct mirror *
3615 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3619 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3620 if (uuid_equals(uuid, &m->uuid)) {
3628 bridge_configure_mirrors(struct bridge *br)
3630 const struct ovsdb_datum *mc;
3631 unsigned long *flood_vlans;
3632 struct mirror *m, *next;
3635 /* Get rid of deleted mirrors. */
3636 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3637 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3638 union ovsdb_atom atom;
3640 atom.uuid = m->uuid;
3641 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3646 /* Add new mirrors and reconfigure existing ones. */
3647 for (i = 0; i < br->cfg->n_mirrors; i++) {
3648 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3649 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3651 m = mirror_create(br, cfg);
3654 if (!mirror_configure(m)) {
3659 /* Update flooded vlans (for RSPAN). */
3660 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3661 br->cfg->n_flood_vlans);
3662 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3663 bitmap_free(flood_vlans);
3666 static struct mirror *
3667 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3671 m = xzalloc(sizeof *m);
3672 m->uuid = cfg->header_.uuid;
3673 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3675 m->name = xstrdup(cfg->name);
3681 mirror_destroy(struct mirror *m)
3684 struct bridge *br = m->bridge;
3687 ofproto_mirror_unregister(br->ofproto, m);
3690 hmap_remove(&br->mirrors, &m->hmap_node);
3697 mirror_collect_ports(struct mirror *m,
3698 struct ovsrec_port **in_ports, int n_in_ports,
3699 void ***out_portsp, size_t *n_out_portsp)
3701 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3702 size_t n_out_ports = 0;
3705 for (i = 0; i < n_in_ports; i++) {
3706 const char *name = in_ports[i]->name;
3707 struct port *port = port_lookup(m->bridge, name);
3709 out_ports[n_out_ports++] = port;
3711 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3712 "port %s", m->bridge->name, m->name, name);
3715 *out_portsp = out_ports;
3716 *n_out_portsp = n_out_ports;
3720 mirror_configure(struct mirror *m)
3722 const struct ovsrec_mirror *cfg = m->cfg;
3723 struct ofproto_mirror_settings s;
3726 if (strcmp(cfg->name, m->name)) {
3728 m->name = xstrdup(cfg->name);
3732 /* Get output port or VLAN. */
3733 if (cfg->output_port) {
3734 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3735 if (!s.out_bundle) {
3736 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3737 m->bridge->name, m->name);
3740 s.out_vlan = UINT16_MAX;
3742 if (cfg->output_vlan) {
3743 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3744 "output vlan; ignoring output vlan",
3745 m->bridge->name, m->name);
3747 } else if (cfg->output_vlan) {
3748 /* The database should prevent invalid VLAN values. */
3749 s.out_bundle = NULL;
3750 s.out_vlan = *cfg->output_vlan;
3752 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3753 m->bridge->name, m->name);
3757 /* Get port selection. */
3758 if (cfg->select_all) {
3759 size_t n_ports = hmap_count(&m->bridge->ports);
3760 void **ports = xmalloc(n_ports * sizeof *ports);
3765 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3775 /* Get ports, dropping ports that don't exist.
3776 * The IDL ensures that there are no duplicates. */
3777 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3778 &s.srcs, &s.n_srcs);
3779 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3780 &s.dsts, &s.n_dsts);
3783 /* Get VLAN selection. */
3784 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3787 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3790 if (s.srcs != s.dsts) {
3799 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3801 * This is deprecated. It is only for compatibility with broken device drivers
3802 * in old versions of Linux that do not properly support VLANs when VLAN
3803 * devices are not used. When broken device drivers are no longer in
3804 * widespread use, we will delete these interfaces. */
3806 static struct ovsrec_port **recs;
3807 static size_t n_recs, allocated_recs;
3809 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3810 * splinters are reconfigured. */
3812 register_rec(struct ovsrec_port *rec)
3814 if (n_recs >= allocated_recs) {
3815 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3817 recs[n_recs++] = rec;
3820 /* Frees all of the ports registered with register_reg(). */
3822 free_registered_recs(void)
3826 for (i = 0; i < n_recs; i++) {
3827 struct ovsrec_port *port = recs[i];
3830 for (j = 0; j < port->n_interfaces; j++) {
3831 struct ovsrec_interface *iface = port->interfaces[j];
3836 smap_destroy(&port->other_config);
3837 free(port->interfaces);
3845 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3848 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3850 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3854 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3857 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3858 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3859 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3862 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3863 * use, returns NULL.
3865 * Updates 'vlan_splinters_enabled_anywhere'. */
3866 static unsigned long int *
3867 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3869 unsigned long int *splinter_vlans;
3870 struct sset splinter_ifaces;
3871 const char *real_dev_name;
3872 struct shash *real_devs;
3873 struct shash_node *node;
3877 /* Free space allocated for synthesized ports and interfaces, since we're
3878 * in the process of reconstructing all of them. */
3879 free_registered_recs();
3881 splinter_vlans = bitmap_allocate(4096);
3882 sset_init(&splinter_ifaces);
3883 vlan_splinters_enabled_anywhere = false;
3884 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3885 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3888 for (j = 0; j < br_cfg->n_ports; j++) {
3889 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3892 for (k = 0; k < port_cfg->n_interfaces; k++) {
3893 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3895 if (vlan_splinters_is_enabled(iface_cfg)) {
3896 vlan_splinters_enabled_anywhere = true;
3897 sset_add(&splinter_ifaces, iface_cfg->name);
3898 vlan_bitmap_from_array__(port_cfg->trunks,
3904 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3905 bitmap_set1(splinter_vlans, *port_cfg->tag);
3910 if (!vlan_splinters_enabled_anywhere) {
3911 free(splinter_vlans);
3912 sset_destroy(&splinter_ifaces);
3916 HMAP_FOR_EACH (br, node, &all_bridges) {
3918 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3922 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3923 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3924 * device to be created for it. */
3925 bitmap_set0(splinter_vlans, 0);
3926 bitmap_set0(splinter_vlans, 4095);
3928 /* Delete all VLAN devices that we don't need. */
3930 real_devs = vlandev_get_real_devs();
3931 SHASH_FOR_EACH (node, real_devs) {
3932 const struct vlan_real_dev *real_dev = node->data;
3933 const struct vlan_dev *vlan_dev;
3934 bool real_dev_has_splinters;
3936 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3938 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3939 if (!real_dev_has_splinters
3940 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3941 struct netdev *netdev;
3943 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3944 if (!netdev_get_in4(netdev, NULL, NULL) ||
3945 !netdev_get_in6(netdev, NULL)) {
3946 vlandev_del(vlan_dev->name);
3948 /* It has an IP address configured, so we don't own
3949 * it. Don't delete it. */
3951 netdev_close(netdev);
3958 /* Add all VLAN devices that we need. */
3959 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3962 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3963 if (!vlandev_get_name(real_dev_name, vid)) {
3964 vlandev_add(real_dev_name, vid);
3971 sset_destroy(&splinter_ifaces);
3973 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
3974 free(splinter_vlans);
3977 return splinter_vlans;
3980 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
3983 configure_splinter_port(struct port *port)
3985 struct ofproto *ofproto = port->bridge->ofproto;
3986 uint16_t realdev_ofp_port;
3987 const char *realdev_name;
3988 struct iface *vlandev, *realdev;
3990 ofproto_bundle_unregister(port->bridge->ofproto, port);
3992 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
3995 realdev_name = smap_get(&port->cfg->other_config, "realdev");
3996 realdev = iface_lookup(port->bridge, realdev_name);
3997 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
3999 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4003 static struct ovsrec_port *
4004 synthesize_splinter_port(const char *real_dev_name,
4005 const char *vlan_dev_name, int vid)
4007 struct ovsrec_interface *iface;
4008 struct ovsrec_port *port;
4010 iface = xmalloc(sizeof *iface);
4011 ovsrec_interface_init(iface);
4012 iface->name = xstrdup(vlan_dev_name);
4013 iface->type = "system";
4015 port = xmalloc(sizeof *port);
4016 ovsrec_port_init(port);
4017 port->interfaces = xmemdup(&iface, sizeof iface);
4018 port->n_interfaces = 1;
4019 port->name = xstrdup(vlan_dev_name);
4020 port->vlan_mode = "splinter";
4021 port->tag = xmalloc(sizeof *port->tag);
4024 smap_add(&port->other_config, "realdev", real_dev_name);
4030 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4031 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4032 * 1-bit in the 'splinter_vlans' bitmap. */
4034 add_vlan_splinter_ports(struct bridge *br,
4035 const unsigned long int *splinter_vlans,
4036 struct shash *ports)
4040 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4041 * we're modifying 'ports'. */
4042 for (i = 0; i < br->cfg->n_ports; i++) {
4043 const char *name = br->cfg->ports[i]->name;
4044 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4047 for (j = 0; j < port_cfg->n_interfaces; j++) {
4048 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4050 if (vlan_splinters_is_enabled(iface_cfg)) {
4051 const char *real_dev_name;
4054 real_dev_name = iface_cfg->name;
4055 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4056 const char *vlan_dev_name;
4058 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4060 && !shash_find(ports, vlan_dev_name)) {
4061 shash_add(ports, vlan_dev_name,
4062 synthesize_splinter_port(
4063 real_dev_name, vlan_dev_name, vid));
4072 mirror_refresh_stats(struct mirror *m)
4074 struct ofproto *ofproto = m->bridge->ofproto;
4075 uint64_t tx_packets, tx_bytes;
4078 size_t stat_cnt = 0;
4080 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4081 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4085 if (tx_packets != UINT64_MAX) {
4086 keys[stat_cnt] = "tx_packets";
4087 values[stat_cnt] = tx_packets;
4090 if (tx_bytes != UINT64_MAX) {
4091 keys[stat_cnt] = "tx_bytes";
4092 values[stat_cnt] = tx_bytes;
4096 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);