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 uint32_t *bond_stable_ids);
234 static bool port_is_synthetic(const struct port *);
236 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
237 static void run_system_stats(void);
239 static void bridge_configure_mirrors(struct bridge *);
240 static struct mirror *mirror_create(struct bridge *,
241 const struct ovsrec_mirror *);
242 static void mirror_destroy(struct mirror *);
243 static bool mirror_configure(struct mirror *);
244 static void mirror_refresh_stats(struct mirror *);
246 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
247 static bool iface_create(struct bridge *, struct if_cfg *, int ofp_port);
248 static bool iface_is_internal(const struct ovsrec_interface *iface,
249 const struct ovsrec_bridge *br);
250 static const char *iface_get_type(const struct ovsrec_interface *,
251 const struct ovsrec_bridge *);
252 static void iface_destroy(struct iface *);
253 static struct iface *iface_lookup(const struct bridge *, const char *name);
254 static struct iface *iface_find(const char *name);
255 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
256 static struct iface *iface_from_ofp_port(const struct bridge *,
258 static void iface_set_mac(struct iface *);
259 static void iface_set_ofport(const struct ovsrec_interface *, int64_t ofport);
260 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
261 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
262 static void iface_configure_cfm(struct iface *);
263 static void iface_refresh_cfm_stats(struct iface *);
264 static void iface_refresh_stats(struct iface *);
265 static void iface_refresh_status(struct iface *);
266 static bool iface_is_synthetic(const struct iface *);
267 static int64_t iface_pick_ofport(const struct ovsrec_interface *);
269 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
271 * This is deprecated. It is only for compatibility with broken device drivers
272 * in old versions of Linux that do not properly support VLANs when VLAN
273 * devices are not used. When broken device drivers are no longer in
274 * widespread use, we will delete these interfaces. */
276 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
277 static bool vlan_splinters_enabled_anywhere;
279 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
280 static unsigned long int *collect_splinter_vlans(
281 const struct ovsrec_open_vswitch *);
282 static void configure_splinter_port(struct port *);
283 static void add_vlan_splinter_ports(struct bridge *,
284 const unsigned long int *splinter_vlans,
285 struct shash *ports);
288 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
290 struct shash iface_hints;
291 static bool initialized = false;
298 shash_init(&iface_hints);
301 for (i = 0; i < cfg->n_bridges; i++) {
302 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
305 for (j = 0; j < br_cfg->n_ports; j++) {
306 struct ovsrec_port *port_cfg = br_cfg->ports[j];
309 for (k = 0; k < port_cfg->n_interfaces; k++) {
310 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
311 struct iface_hint *iface_hint;
313 iface_hint = xmalloc(sizeof *iface_hint);
314 iface_hint->br_name = br_cfg->name;
315 iface_hint->br_type = br_cfg->datapath_type;
316 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
318 shash_add(&iface_hints, if_cfg->name, iface_hint);
324 ofproto_init(&iface_hints);
326 shash_destroy_free_data(&iface_hints);
330 /* Public functions. */
332 /* Initializes the bridge module, configuring it to obtain its configuration
333 * from an OVSDB server accessed over 'remote', which should be a string in a
334 * form acceptable to ovsdb_idl_create(). */
336 bridge_init(const char *remote)
338 /* Create connection to database. */
339 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
340 idl_seqno = ovsdb_idl_get_seqno(idl);
341 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
342 ovsdb_idl_verify_write_only(idl);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
345 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
353 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
354 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
357 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
358 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
359 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
377 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
380 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
381 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
382 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
384 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
386 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
389 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
391 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
397 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
398 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
401 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
403 /* Register unixctl commands. */
404 unixctl_command_register("qos/show", "interface", 1, 1,
405 qos_unixctl_show, NULL);
406 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
407 bridge_unixctl_dump_flows, NULL);
408 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
409 bridge_unixctl_reconnect, NULL);
419 struct bridge *br, *next_br;
421 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
424 ovsdb_idl_destroy(idl);
427 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
428 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
429 * responsible for freeing '*managersp' (with free()).
431 * You may be asking yourself "why does ovs-vswitchd care?", because
432 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
433 * should not be and in fact is not directly involved in that. But
434 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
435 * it has to tell in-band control where the managers are to enable that.
436 * (Thus, only managers connected in-band are collected.)
439 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
440 struct sockaddr_in **managersp, size_t *n_managersp)
442 struct sockaddr_in *managers = NULL;
443 size_t n_managers = 0;
447 /* Collect all of the potential targets from the "targets" columns of the
448 * rows pointed to by "manager_options", excluding any that are
451 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
452 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
454 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
455 sset_find_and_delete(&targets, m->target);
457 sset_add(&targets, m->target);
461 /* Now extract the targets' IP addresses. */
462 if (!sset_is_empty(&targets)) {
465 managers = xmalloc(sset_count(&targets) * sizeof *managers);
466 SSET_FOR_EACH (target, &targets) {
467 struct sockaddr_in *sin = &managers[n_managers];
469 if (stream_parse_target_with_default_ports(target,
477 sset_destroy(&targets);
479 *managersp = managers;
480 *n_managersp = n_managers;
484 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
486 unsigned long int *splinter_vlans;
489 COVERAGE_INC(bridge_reconfigure);
491 ovs_assert(!reconfiguring);
492 reconfiguring = true;
494 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
495 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
496 * configuration otherwise.
498 * This is mostly an update to bridge data structures. Nothing is pushed
499 * down to ofproto or lower layers. */
500 add_del_bridges(ovs_cfg);
501 splinter_vlans = collect_splinter_vlans(ovs_cfg);
502 HMAP_FOR_EACH (br, node, &all_bridges) {
503 bridge_add_del_ports(br, splinter_vlans);
505 free(splinter_vlans);
507 /* Delete datapaths that are no longer configured, and create ones which
508 * don't exist but should. */
509 bridge_update_ofprotos();
511 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
512 HMAP_FOR_EACH (br, node, &all_bridges) {
513 bridge_refresh_ofp_port(br);
516 /* Clear database records for "if_cfg"s which haven't been instantiated. */
517 HMAP_FOR_EACH (br, node, &all_bridges) {
518 struct if_cfg *if_cfg;
520 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
521 iface_clear_db_record(if_cfg->cfg);
525 reconfigure_system_stats(ovs_cfg);
529 bridge_reconfigure_ofp(void)
531 long long int deadline;
535 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
537 /* The kernel will reject any attempt to add a given port to a datapath if
538 * that port already belongs to a different datapath, so we must do all
539 * port deletions before any port additions. */
540 HMAP_FOR_EACH (br, node, &all_bridges) {
541 struct ofpp_garbage *garbage, *next;
543 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
544 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
545 * internal datastructures may not be consistent. Eventually, when
546 * port additions and deletions are cheaper, these calls should be
549 ofproto_port_del(br->ofproto, garbage->ofp_port);
550 list_remove(&garbage->list_node);
554 if (time_msec() >= deadline) {
561 HMAP_FOR_EACH (br, node, &all_bridges) {
562 struct if_cfg *if_cfg, *next;
564 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
565 iface_create(br, if_cfg, -1);
567 if (time_msec() >= deadline) {
577 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
579 struct sockaddr_in *managers;
580 int sflow_bridge_number;
585 ovs_assert(reconfiguring);
586 done = bridge_reconfigure_ofp();
588 /* Complete the configuration. */
589 sflow_bridge_number = 0;
590 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
591 HMAP_FOR_EACH (br, node, &all_bridges) {
594 /* We need the datapath ID early to allow LACP ports to use it as the
595 * default system ID. */
596 bridge_configure_datapath_id(br);
598 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
601 port_configure(port);
603 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
604 iface_configure_cfm(iface);
605 iface_configure_qos(iface, port->cfg->qos);
606 iface_set_mac(iface);
609 bridge_configure_mirrors(br);
610 bridge_configure_flow_eviction_threshold(br);
611 bridge_configure_forward_bpdu(br);
612 bridge_configure_mac_table(br);
613 bridge_configure_remotes(br, managers, n_managers);
614 bridge_configure_netflow(br);
615 bridge_configure_sflow(br, &sflow_bridge_number);
616 bridge_configure_stp(br);
617 bridge_configure_tables(br);
618 bridge_configure_dp_desc(br);
625 /* Delete ofprotos which aren't configured or have the wrong type. Create
626 * ofprotos which don't exist but need to. */
628 bridge_update_ofprotos(void)
630 struct bridge *br, *next;
635 /* Delete ofprotos with no bridge or with the wrong type. */
638 ofproto_enumerate_types(&types);
639 SSET_FOR_EACH (type, &types) {
642 ofproto_enumerate_names(type, &names);
643 SSET_FOR_EACH (name, &names) {
644 br = bridge_lookup(name);
645 if (!br || strcmp(type, br->type)) {
646 ofproto_delete(name, type);
650 sset_destroy(&names);
651 sset_destroy(&types);
653 /* Add ofprotos for bridges which don't have one yet. */
654 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
662 /* Remove ports from any datapath with the same name as 'br'. If we
663 * don't do this, creating 'br''s ofproto will fail because a port with
664 * the same name as its local port already exists. */
665 HMAP_FOR_EACH (br2, node, &all_bridges) {
666 struct ofproto_port ofproto_port;
672 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
674 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
676 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
679 ofproto_port_destroy(&ofproto_port);
683 error = ofproto_create(br->name, br->type, &br->ofproto);
685 VLOG_ERR("failed to create bridge %s: %s", br->name,
693 port_configure(struct port *port)
695 const struct ovsrec_port *cfg = port->cfg;
696 struct bond_settings bond_settings;
697 struct lacp_settings lacp_settings;
698 struct ofproto_bundle_settings s;
701 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
702 configure_splinter_port(port);
711 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
712 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
713 s.slaves[s.n_slaves++] = iface->ofp_port;
718 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
722 /* Get VLAN trunks. */
725 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
729 if (cfg->vlan_mode) {
730 if (!strcmp(cfg->vlan_mode, "access")) {
731 s.vlan_mode = PORT_VLAN_ACCESS;
732 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
733 s.vlan_mode = PORT_VLAN_TRUNK;
734 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
735 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
736 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
737 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
739 /* This "can't happen" because ovsdb-server should prevent it. */
740 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
741 s.vlan_mode = PORT_VLAN_TRUNK;
745 s.vlan_mode = PORT_VLAN_ACCESS;
747 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
751 s.vlan_mode = PORT_VLAN_TRUNK;
754 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
757 /* Get LACP settings. */
758 s.lacp = port_configure_lacp(port, &lacp_settings);
762 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
763 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
764 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
767 s.lacp_slaves = NULL;
770 /* Get bond settings. */
771 if (s.n_slaves > 1) {
772 s.bond = &bond_settings;
773 s.bond_stable_ids = xmalloc(s.n_slaves * sizeof *s.bond_stable_ids);
774 port_configure_bond(port, &bond_settings, s.bond_stable_ids);
777 s.bond_stable_ids = NULL;
779 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
780 netdev_set_miimon_interval(iface->netdev, 0);
785 ofproto_bundle_register(port->bridge->ofproto, port, &s);
791 free(s.bond_stable_ids);
794 /* Pick local port hardware address and datapath ID for 'br'. */
796 bridge_configure_datapath_id(struct bridge *br)
798 uint8_t ea[ETH_ADDR_LEN];
800 struct iface *local_iface;
801 struct iface *hw_addr_iface;
804 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
805 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
807 int error = netdev_set_etheraddr(local_iface->netdev, ea);
809 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
810 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
811 "Ethernet address: %s",
812 br->name, strerror(error));
815 memcpy(br->ea, ea, ETH_ADDR_LEN);
817 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
818 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
819 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
820 ofproto_set_datapath_id(br->ofproto, dpid);
823 dpid_string = xasprintf("%016"PRIx64, dpid);
824 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
828 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
831 bridge_get_allowed_versions(struct bridge *br)
833 if (!br->cfg->n_protocols)
836 return ofputil_versions_from_strings(br->cfg->protocols,
837 br->cfg->n_protocols);
840 /* Set NetFlow configuration on 'br'. */
842 bridge_configure_netflow(struct bridge *br)
844 struct ovsrec_netflow *cfg = br->cfg->netflow;
845 struct netflow_options opts;
848 ofproto_set_netflow(br->ofproto, NULL);
852 memset(&opts, 0, sizeof opts);
854 /* Get default NetFlow configuration from datapath.
855 * Apply overrides from 'cfg'. */
856 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
857 if (cfg->engine_type) {
858 opts.engine_type = *cfg->engine_type;
860 if (cfg->engine_id) {
861 opts.engine_id = *cfg->engine_id;
864 /* Configure active timeout interval. */
865 opts.active_timeout = cfg->active_timeout;
866 if (!opts.active_timeout) {
867 opts.active_timeout = -1;
868 } else if (opts.active_timeout < 0) {
869 VLOG_WARN("bridge %s: active timeout interval set to negative "
870 "value, using default instead (%d seconds)", br->name,
871 NF_ACTIVE_TIMEOUT_DEFAULT);
872 opts.active_timeout = -1;
875 /* Add engine ID to interface number to disambiguate bridgs? */
876 opts.add_id_to_iface = cfg->add_id_to_interface;
877 if (opts.add_id_to_iface) {
878 if (opts.engine_id > 0x7f) {
879 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
880 "another vswitch, choose an engine id less than 128",
883 if (hmap_count(&br->ports) > 508) {
884 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
885 "another port when more than 508 ports are used",
891 sset_init(&opts.collectors);
892 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
895 if (ofproto_set_netflow(br->ofproto, &opts)) {
896 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
898 sset_destroy(&opts.collectors);
901 /* Set sFlow configuration on 'br'. */
903 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
905 const struct ovsrec_sflow *cfg = br->cfg->sflow;
906 struct ovsrec_controller **controllers;
907 struct ofproto_sflow_options oso;
908 size_t n_controllers;
912 ofproto_set_sflow(br->ofproto, NULL);
916 memset(&oso, 0, sizeof oso);
918 sset_init(&oso.targets);
919 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
921 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
923 oso.sampling_rate = *cfg->sampling;
926 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
928 oso.polling_interval = *cfg->polling;
931 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
933 oso.header_len = *cfg->header;
936 oso.sub_id = (*sflow_bridge_number)++;
937 oso.agent_device = cfg->agent;
939 oso.control_ip = NULL;
940 n_controllers = bridge_get_controllers(br, &controllers);
941 for (i = 0; i < n_controllers; i++) {
942 if (controllers[i]->local_ip) {
943 oso.control_ip = controllers[i]->local_ip;
947 ofproto_set_sflow(br->ofproto, &oso);
949 sset_destroy(&oso.targets);
953 port_configure_stp(const struct ofproto *ofproto, struct port *port,
954 struct ofproto_port_stp_settings *port_s,
955 int *port_num_counter, unsigned long *port_num_bitmap)
957 const char *config_str;
960 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
961 port_s->enable = false;
964 port_s->enable = true;
967 /* STP over bonds is not supported. */
968 if (!list_is_singleton(&port->ifaces)) {
969 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
971 port_s->enable = false;
975 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
977 /* Internal ports shouldn't participate in spanning tree, so
979 if (!strcmp(iface->type, "internal")) {
980 VLOG_DBG("port %s: disable STP on internal ports", port->name);
981 port_s->enable = false;
985 /* STP on mirror output ports is not supported. */
986 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
987 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
988 port_s->enable = false;
992 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
994 unsigned long int port_num = strtoul(config_str, NULL, 0);
995 int port_idx = port_num - 1;
997 if (port_num < 1 || port_num > STP_MAX_PORTS) {
998 VLOG_ERR("port %s: invalid stp-port-num", port->name);
999 port_s->enable = false;
1003 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1004 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1005 port->name, port_num);
1006 port_s->enable = false;
1009 bitmap_set1(port_num_bitmap, port_idx);
1010 port_s->port_num = port_idx;
1012 if (*port_num_counter >= STP_MAX_PORTS) {
1013 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1014 port_s->enable = false;
1018 port_s->port_num = (*port_num_counter)++;
1021 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1023 port_s->path_cost = strtoul(config_str, NULL, 10);
1025 enum netdev_features current;
1028 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1029 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1030 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1033 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1035 port_s->priority = strtoul(config_str, NULL, 0);
1037 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1041 /* Set spanning tree configuration on 'br'. */
1043 bridge_configure_stp(struct bridge *br)
1045 if (!br->cfg->stp_enable) {
1046 ofproto_set_stp(br->ofproto, NULL);
1048 struct ofproto_stp_settings br_s;
1049 const char *config_str;
1051 int port_num_counter;
1052 unsigned long *port_num_bitmap;
1054 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1056 uint8_t ea[ETH_ADDR_LEN];
1058 if (eth_addr_from_string(config_str, ea)) {
1059 br_s.system_id = eth_addr_to_uint64(ea);
1061 br_s.system_id = eth_addr_to_uint64(br->ea);
1062 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1063 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1066 br_s.system_id = eth_addr_to_uint64(br->ea);
1069 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1071 br_s.priority = strtoul(config_str, NULL, 0);
1073 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1076 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1078 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1080 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1083 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1085 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1087 br_s.max_age = STP_DEFAULT_MAX_AGE;
1090 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1092 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1094 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1097 /* Configure STP on the bridge. */
1098 if (ofproto_set_stp(br->ofproto, &br_s)) {
1099 VLOG_ERR("bridge %s: could not enable STP", br->name);
1103 /* Users must either set the port number with the "stp-port-num"
1104 * configuration on all ports or none. If manual configuration
1105 * is not done, then we allocate them sequentially. */
1106 port_num_counter = 0;
1107 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1108 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1109 struct ofproto_port_stp_settings port_s;
1110 struct iface *iface;
1112 port_configure_stp(br->ofproto, port, &port_s,
1113 &port_num_counter, port_num_bitmap);
1115 /* As bonds are not supported, just apply configuration to
1116 * all interfaces. */
1117 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1118 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1120 VLOG_ERR("port %s: could not enable STP", port->name);
1126 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1127 && port_num_counter) {
1128 VLOG_ERR("bridge %s: must manually configure all STP port "
1129 "IDs or none, disabling", br->name);
1130 ofproto_set_stp(br->ofproto, NULL);
1132 bitmap_free(port_num_bitmap);
1137 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1139 const struct port *port = port_lookup(br, name);
1140 return port && port_is_bond_fake_iface(port);
1144 port_is_bond_fake_iface(const struct port *port)
1146 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1150 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1152 struct bridge *br, *next;
1153 struct shash new_br;
1156 /* Collect new bridges' names and types. */
1157 shash_init(&new_br);
1158 for (i = 0; i < cfg->n_bridges; i++) {
1159 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1160 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1162 if (strchr(br_cfg->name, '/')) {
1163 /* Prevent remote ovsdb-server users from accessing arbitrary
1164 * directories, e.g. consider a bridge named "../../../etc/". */
1165 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1167 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1168 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1172 /* Get rid of deleted bridges or those whose types have changed.
1173 * Update 'cfg' of bridges that still exist. */
1174 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1175 br->cfg = shash_find_data(&new_br, br->name);
1176 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1177 br->cfg->datapath_type))) {
1182 /* Add new bridges. */
1183 for (i = 0; i < cfg->n_bridges; i++) {
1184 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1185 struct bridge *br = bridge_lookup(br_cfg->name);
1187 bridge_create(br_cfg);
1191 shash_destroy(&new_br);
1195 iface_set_ofp_port(struct iface *iface, int ofp_port)
1197 struct bridge *br = iface->port->bridge;
1199 ovs_assert(iface->ofp_port < 0 && ofp_port >= 0);
1200 iface->ofp_port = ofp_port;
1201 hmap_insert(&br->ifaces, &iface->ofp_port_node, hash_int(ofp_port, 0));
1202 iface_set_ofport(iface->cfg, ofp_port);
1205 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1206 * Returns 0 if successful, otherwise a positive errno value. */
1208 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1209 struct netdev *netdev)
1213 error = netdev_set_config(netdev, &iface_cfg->options);
1215 VLOG_WARN("could not configure network device %s (%s)",
1216 iface_cfg->name, strerror(error));
1221 /* This function determines whether 'ofproto_port', which is attached to
1222 * br->ofproto's datapath, is one that we want in 'br'.
1224 * If it is, it returns true, after creating an iface (if necessary),
1225 * configuring the iface's netdev according to the iface's options, and setting
1226 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1228 * If, on the other hand, 'port' should be removed, it returns false. The
1229 * caller should later detach the port from br->ofproto. */
1231 bridge_refresh_one_ofp_port(struct bridge *br,
1232 const struct ofproto_port *ofproto_port)
1234 const char *name = ofproto_port->name;
1235 const char *type = ofproto_port->type;
1236 uint16_t ofp_port = ofproto_port->ofp_port;
1238 struct iface *iface = iface_lookup(br, name);
1240 /* Check that the name-to-number mapping is one-to-one. */
1241 if (iface->ofp_port >= 0) {
1242 VLOG_WARN("bridge %s: interface %s reported twice",
1245 } else if (iface_from_ofp_port(br, ofp_port)) {
1246 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1247 br->name, ofp_port);
1251 /* There's a configured interface named 'name'. */
1252 if (strcmp(type, iface->type)
1253 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1254 /* It's the wrong type, or it's the right type but can't be
1255 * configured as the user requested, so we must destroy it. */
1258 /* It's the right type and configured correctly. Keep it. */
1259 iface_set_ofp_port(iface, ofp_port);
1262 } else if (bridge_has_bond_fake_iface(br, name)
1263 && !strcmp(type, "internal")) {
1264 /* It's a bond fake iface. Keep it. */
1267 /* There's no configured interface named 'name', but there might be an
1268 * interface of that name queued to be created.
1270 * If there is, and it has the correct type, then try to configure it
1271 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1272 * delete it and later try to re-add it. */
1273 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1275 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1276 && iface_create(br, if_cfg, ofp_port));
1280 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1281 * consistent with the ofp_ports in "br->ofproto". */
1283 bridge_refresh_ofp_port(struct bridge *br)
1285 struct ofproto_port_dump dump;
1286 struct ofproto_port ofproto_port;
1287 struct port *port, *port_next;
1289 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1290 hmap_clear(&br->ifaces);
1291 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1292 struct iface *iface;
1294 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1295 iface->ofp_port = -1;
1299 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1300 * already exist in the datapath and promote them to full fledged "struct
1301 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1302 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1303 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1304 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1305 garbage->ofp_port = ofproto_port.ofp_port;
1306 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1310 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1311 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1312 * later they can be added to ofproto. */
1313 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1314 struct iface *iface, *iface_next;
1316 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1317 if (iface->ofp_port < 0) {
1318 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1319 iface_destroy(iface);
1323 if (list_is_empty(&port->ifaces)) {
1329 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1330 * is negative, adds the network device to br->ofproto and stores the OpenFlow
1331 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1334 * If successful, returns 0 and stores the network device in '*netdevp'. On
1335 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1337 iface_do_create(const struct bridge *br,
1338 const struct if_cfg *if_cfg,
1339 int *ofp_portp, struct netdev **netdevp)
1341 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1342 const struct ovsrec_port *port_cfg = if_cfg->parent;
1343 struct netdev *netdev;
1346 error = netdev_open(iface_cfg->name,
1347 iface_get_type(iface_cfg, br->cfg), &netdev);
1349 VLOG_WARN("could not open network device %s (%s)",
1350 iface_cfg->name, strerror(error));
1354 error = iface_set_netdev_config(iface_cfg, netdev);
1359 if (*ofp_portp < 0) {
1360 uint16_t ofp_port = if_cfg->ofport;
1362 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1366 *ofp_portp = ofp_port;
1368 VLOG_INFO("bridge %s: added interface %s on port %d",
1369 br->name, iface_cfg->name, *ofp_portp);
1371 VLOG_DBG("bridge %s: interface %s is on port %d",
1372 br->name, iface_cfg->name, *ofp_portp);
1375 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1376 || iface_is_internal(iface_cfg, br->cfg)) {
1377 netdev_turn_flags_on(netdev, NETDEV_UP, true);
1385 netdev_close(netdev);
1389 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1390 * port number 'ofp_port'. If ofp_port is negative, an OpenFlow port is
1391 * automatically allocated for the iface. Takes ownership of and
1392 * deallocates 'if_cfg'.
1394 * Return true if an iface is successfully created, false otherwise. */
1396 iface_create(struct bridge *br, struct if_cfg *if_cfg, int ofp_port)
1398 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1399 const struct ovsrec_port *port_cfg = if_cfg->parent;
1401 struct netdev *netdev;
1402 struct iface *iface;
1407 /* Do the bits that can fail up front.
1409 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1410 * internal datastructures may not be consistent. Eventually, when port
1411 * additions and deletions are cheaper, these calls should be removed. */
1413 ovs_assert(!iface_lookup(br, iface_cfg->name));
1414 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1417 iface_set_ofport(iface_cfg, -1);
1418 iface_clear_db_record(iface_cfg);
1423 /* Get or create the port structure. */
1424 port = port_lookup(br, port_cfg->name);
1426 port = port_create(br, port_cfg);
1429 /* Create the iface structure. */
1430 iface = xzalloc(sizeof *iface);
1431 list_push_back(&port->ifaces, &iface->port_elem);
1432 hmap_insert(&br->iface_by_name, &iface->name_node,
1433 hash_string(iface_cfg->name, 0));
1435 iface->name = xstrdup(iface_cfg->name);
1436 iface->ofp_port = -1;
1437 iface->netdev = netdev;
1438 iface->type = iface_get_type(iface_cfg, br->cfg);
1439 iface->cfg = iface_cfg;
1441 iface_set_ofp_port(iface, ofp_port);
1443 /* Populate initial status in database. */
1444 iface_refresh_stats(iface);
1445 iface_refresh_status(iface);
1447 /* Add bond fake iface if necessary. */
1448 if (port_is_bond_fake_iface(port)) {
1449 struct ofproto_port ofproto_port;
1451 if (ofproto_port_query_by_name(br->ofproto, port->name,
1453 struct netdev *netdev;
1456 error = netdev_open(port->name, "internal", &netdev);
1458 uint16_t fake_ofp_port = if_cfg->ofport;
1460 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1461 netdev_close(netdev);
1463 VLOG_WARN("could not open network device %s (%s)",
1464 port->name, strerror(error));
1467 /* Already exists, nothing to do. */
1468 ofproto_port_destroy(&ofproto_port);
1473 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1479 /* Set Flow eviction threshold */
1481 bridge_configure_flow_eviction_threshold(struct bridge *br)
1483 const char *threshold_str;
1486 threshold_str = smap_get(&br->cfg->other_config,
1487 "flow-eviction-threshold");
1488 if (threshold_str) {
1489 threshold = strtoul(threshold_str, NULL, 10);
1491 threshold = OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT;
1493 ofproto_set_flow_eviction_threshold(br->ofproto, threshold);
1496 /* Set forward BPDU option. */
1498 bridge_configure_forward_bpdu(struct bridge *br)
1500 ofproto_set_forward_bpdu(br->ofproto,
1501 smap_get_bool(&br->cfg->other_config,
1506 /* Set MAC learning table configuration for 'br'. */
1508 bridge_configure_mac_table(struct bridge *br)
1510 const char *idle_time_str;
1513 const char *mac_table_size_str;
1516 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1517 idle_time = (idle_time_str && atoi(idle_time_str)
1518 ? atoi(idle_time_str)
1519 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1521 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1522 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1523 ? atoi(mac_table_size_str)
1526 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1530 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1531 struct iface **hw_addr_iface)
1533 struct hmapx mirror_output_ports;
1536 bool found_addr = false;
1540 *hw_addr_iface = NULL;
1542 /* Did the user request a particular MAC? */
1543 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1544 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1545 if (eth_addr_is_multicast(ea)) {
1546 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1547 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1548 } else if (eth_addr_is_zero(ea)) {
1549 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1555 /* Mirror output ports don't participate in picking the local hardware
1556 * address. ofproto can't help us find out whether a given port is a
1557 * mirror output because we haven't configured mirrors yet, so we need to
1558 * accumulate them ourselves. */
1559 hmapx_init(&mirror_output_ports);
1560 for (i = 0; i < br->cfg->n_mirrors; i++) {
1561 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1562 if (m->output_port) {
1563 hmapx_add(&mirror_output_ports, m->output_port);
1567 /* Otherwise choose the minimum non-local MAC address among all of the
1569 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1570 uint8_t iface_ea[ETH_ADDR_LEN];
1571 struct iface *candidate;
1572 struct iface *iface;
1574 /* Mirror output ports don't participate. */
1575 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1579 /* Choose the MAC address to represent the port. */
1581 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1582 /* Find the interface with this Ethernet address (if any) so that
1583 * we can provide the correct devname to the caller. */
1584 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1585 uint8_t candidate_ea[ETH_ADDR_LEN];
1586 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1587 && eth_addr_equals(iface_ea, candidate_ea)) {
1592 /* Choose the interface whose MAC address will represent the port.
1593 * The Linux kernel bonding code always chooses the MAC address of
1594 * the first slave added to a bond, and the Fedora networking
1595 * scripts always add slaves to a bond in alphabetical order, so
1596 * for compatibility we choose the interface with the name that is
1597 * first in alphabetical order. */
1598 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1599 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1604 /* The local port doesn't count (since we're trying to choose its
1605 * MAC address anyway). */
1606 if (iface->ofp_port == OFPP_LOCAL) {
1611 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1617 /* Compare against our current choice. */
1618 if (!eth_addr_is_multicast(iface_ea) &&
1619 !eth_addr_is_local(iface_ea) &&
1620 !eth_addr_is_reserved(iface_ea) &&
1621 !eth_addr_is_zero(iface_ea) &&
1622 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1624 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1625 *hw_addr_iface = iface;
1631 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1632 *hw_addr_iface = NULL;
1635 hmapx_destroy(&mirror_output_ports);
1638 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1639 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1640 * an interface on 'br', then that interface must be passed in as
1641 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1642 * 'hw_addr_iface' must be passed in as a null pointer. */
1644 bridge_pick_datapath_id(struct bridge *br,
1645 const uint8_t bridge_ea[ETH_ADDR_LEN],
1646 struct iface *hw_addr_iface)
1649 * The procedure for choosing a bridge MAC address will, in the most
1650 * ordinary case, also choose a unique MAC that we can use as a datapath
1651 * ID. In some special cases, though, multiple bridges will end up with
1652 * the same MAC address. This is OK for the bridges, but it will confuse
1653 * the OpenFlow controller, because each datapath needs a unique datapath
1656 * Datapath IDs must be unique. It is also very desirable that they be
1657 * stable from one run to the next, so that policy set on a datapath
1660 const char *datapath_id;
1663 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1664 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1668 if (!hw_addr_iface) {
1670 * A purely internal bridge, that is, one that has no non-virtual
1671 * network devices on it at all, is difficult because it has no
1672 * natural unique identifier at all.
1674 * When the host is a XenServer, we handle this case by hashing the
1675 * host's UUID with the name of the bridge. Names of bridges are
1676 * persistent across XenServer reboots, although they can be reused if
1677 * an internal network is destroyed and then a new one is later
1678 * created, so this is fairly effective.
1680 * When the host is not a XenServer, we punt by using a random MAC
1681 * address on each run.
1683 const char *host_uuid = xenserver_get_host_uuid();
1685 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1686 dpid = dpid_from_hash(combined, strlen(combined));
1692 return eth_addr_to_uint64(bridge_ea);
1696 dpid_from_hash(const void *data, size_t n)
1698 uint8_t hash[SHA1_DIGEST_SIZE];
1700 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1701 sha1_bytes(data, n, hash);
1702 eth_addr_mark_random(hash);
1703 return eth_addr_to_uint64(hash);
1707 iface_refresh_status(struct iface *iface)
1711 enum netdev_features current;
1715 uint8_t mac[ETH_ADDR_LEN];
1718 if (iface_is_synthetic(iface)) {
1724 if (!netdev_get_status(iface->netdev, &smap)) {
1725 ovsrec_interface_set_status(iface->cfg, &smap);
1727 ovsrec_interface_set_status(iface->cfg, NULL);
1730 smap_destroy(&smap);
1732 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1733 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1735 ovsrec_interface_set_duplex(iface->cfg,
1736 netdev_features_is_full_duplex(current)
1738 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1741 ovsrec_interface_set_duplex(iface->cfg, NULL);
1742 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1745 error = netdev_get_mtu(iface->netdev, &mtu);
1748 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1751 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1754 error = netdev_get_etheraddr(iface->netdev, mac);
1756 char mac_string[32];
1758 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1759 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1761 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1765 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1768 iface_refresh_cfm_stats(struct iface *iface)
1770 const struct ovsrec_interface *cfg = iface->cfg;
1771 int fault, opup, error;
1772 const uint64_t *rmps;
1776 fault = ofproto_port_get_cfm_fault(iface->port->bridge->ofproto,
1779 const char *reasons[CFM_FAULT_N_REASONS];
1780 bool fault_bool = fault;
1784 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1785 int reason = 1 << i;
1786 if (fault & reason) {
1787 reasons[j++] = cfm_fault_reason_to_str(reason);
1791 ovsrec_interface_set_cfm_fault(cfg, &fault_bool, 1);
1792 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1794 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1795 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1798 opup = ofproto_port_get_cfm_opup(iface->port->bridge->ofproto,
1801 ovsrec_interface_set_cfm_remote_opstate(cfg, opup ? "up" : "down");
1803 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1806 error = ofproto_port_get_cfm_remote_mpids(iface->port->bridge->ofproto,
1807 iface->ofp_port, &rmps, &n_rmps);
1809 ovsrec_interface_set_cfm_remote_mpids(cfg, (const int64_t *)rmps,
1812 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1815 health = ofproto_port_get_cfm_health(iface->port->bridge->ofproto,
1818 int64_t cfm_health = health;
1819 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1821 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1826 iface_refresh_stats(struct iface *iface)
1828 #define IFACE_STATS \
1829 IFACE_STAT(rx_packets, "rx_packets") \
1830 IFACE_STAT(tx_packets, "tx_packets") \
1831 IFACE_STAT(rx_bytes, "rx_bytes") \
1832 IFACE_STAT(tx_bytes, "tx_bytes") \
1833 IFACE_STAT(rx_dropped, "rx_dropped") \
1834 IFACE_STAT(tx_dropped, "tx_dropped") \
1835 IFACE_STAT(rx_errors, "rx_errors") \
1836 IFACE_STAT(tx_errors, "tx_errors") \
1837 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1838 IFACE_STAT(rx_over_errors, "rx_over_err") \
1839 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1840 IFACE_STAT(collisions, "collisions")
1842 #define IFACE_STAT(MEMBER, NAME) NAME,
1843 static char *keys[] = { IFACE_STATS };
1845 int64_t values[ARRAY_SIZE(keys)];
1848 struct netdev_stats stats;
1850 if (iface_is_synthetic(iface)) {
1854 /* Intentionally ignore return value, since errors will set 'stats' to
1855 * all-1s, and we will deal with that correctly below. */
1856 netdev_get_stats(iface->netdev, &stats);
1858 /* Copy statistics into values[] array. */
1860 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1863 ovs_assert(i == ARRAY_SIZE(keys));
1865 ovsrec_interface_set_statistics(iface->cfg, keys, values,
1871 br_refresh_stp_status(struct bridge *br)
1873 struct smap smap = SMAP_INITIALIZER(&smap);
1874 struct ofproto *ofproto = br->ofproto;
1875 struct ofproto_stp_status status;
1877 if (ofproto_get_stp_status(ofproto, &status)) {
1881 if (!status.enabled) {
1882 ovsrec_bridge_set_status(br->cfg, NULL);
1886 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1887 STP_ID_ARGS(status.bridge_id));
1888 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1889 STP_ID_ARGS(status.designated_root));
1890 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1892 ovsrec_bridge_set_status(br->cfg, &smap);
1893 smap_destroy(&smap);
1897 port_refresh_stp_status(struct port *port)
1899 struct ofproto *ofproto = port->bridge->ofproto;
1900 struct iface *iface;
1901 struct ofproto_port_stp_status status;
1903 int64_t int_values[3];
1906 if (port_is_synthetic(port)) {
1910 /* STP doesn't currently support bonds. */
1911 if (!list_is_singleton(&port->ifaces)) {
1912 ovsrec_port_set_status(port->cfg, NULL);
1916 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1918 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
1922 if (!status.enabled) {
1923 ovsrec_port_set_status(port->cfg, NULL);
1924 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
1928 /* Set Status column. */
1930 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
1931 smap_add(&smap, "stp_state", stp_state_name(status.state));
1932 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
1933 smap_add(&smap, "stp_role", stp_role_name(status.role));
1934 ovsrec_port_set_status(port->cfg, &smap);
1935 smap_destroy(&smap);
1937 /* Set Statistics column. */
1938 keys[0] = "stp_tx_count";
1939 int_values[0] = status.tx_count;
1940 keys[1] = "stp_rx_count";
1941 int_values[1] = status.rx_count;
1942 keys[2] = "stp_error_count";
1943 int_values[2] = status.error_count;
1945 ovsrec_port_set_statistics(port->cfg, keys, int_values,
1946 ARRAY_SIZE(int_values));
1950 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
1952 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
1956 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
1958 bool enable = enable_system_stats(cfg);
1960 system_stats_enable(enable);
1962 ovsrec_open_vswitch_set_statistics(cfg, NULL);
1967 run_system_stats(void)
1969 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
1972 stats = system_stats_run();
1974 struct ovsdb_idl_txn *txn;
1975 struct ovsdb_datum datum;
1977 txn = ovsdb_idl_txn_create(idl);
1978 ovsdb_datum_from_smap(&datum, stats);
1979 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
1981 ovsdb_idl_txn_commit(txn);
1982 ovsdb_idl_txn_destroy(txn);
1988 static inline const char *
1989 nx_role_to_str(enum nx_role role)
1994 case NX_ROLE_MASTER:
1999 return "*** INVALID ROLE ***";
2004 refresh_controller_status(void)
2008 const struct ovsrec_controller *cfg;
2012 /* Accumulate status for controllers on all bridges. */
2013 HMAP_FOR_EACH (br, node, &all_bridges) {
2014 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2017 /* Update each controller in the database with current status. */
2018 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2019 struct ofproto_controller_info *cinfo =
2020 shash_find_data(&info, cfg->target);
2023 struct smap smap = SMAP_INITIALIZER(&smap);
2024 const char **values = cinfo->pairs.values;
2025 const char **keys = cinfo->pairs.keys;
2028 for (i = 0; i < cinfo->pairs.n; i++) {
2029 smap_add(&smap, keys[i], values[i]);
2032 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2033 ovsrec_controller_set_role(cfg, nx_role_to_str(cinfo->role));
2034 ovsrec_controller_set_status(cfg, &smap);
2035 smap_destroy(&smap);
2037 ovsrec_controller_set_is_connected(cfg, false);
2038 ovsrec_controller_set_role(cfg, NULL);
2039 ovsrec_controller_set_status(cfg, NULL);
2043 ofproto_free_ofproto_controller_info(&info);
2048 * Some information in the database must be kept as up-to-date as possible to
2049 * allow controllers to respond rapidly to network outages. We call these
2050 * statistics "instant" stats.
2052 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2053 * assuming that they've changed. The only means we have to determine whether
2054 * they have changed are:
2056 * - Try to commit changes to the database. If nothing changed, then
2057 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2060 * - instant_stats_run() is called late in the run loop, after anything that
2061 * might change any of the instant stats.
2063 * We use these two facts together to avoid waking the process up every
2064 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2067 /* Minimum interval between writing updates to the instant stats to the
2069 #define INSTANT_INTERVAL_MSEC 100
2071 /* Current instant stats database transaction, NULL if there is no ongoing
2073 static struct ovsdb_idl_txn *instant_txn;
2075 /* Next time (in msec on monotonic clock) at which we will update the instant
2077 static long long int instant_next_txn = LLONG_MIN;
2079 /* True if the run loop has run since we last saw that the instant stats were
2080 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2081 * to refresh the instant stats. */
2082 static bool instant_stats_could_have_changed;
2085 instant_stats_run(void)
2087 enum ovsdb_idl_txn_status status;
2089 instant_stats_could_have_changed = true;
2094 if (time_msec() < instant_next_txn) {
2097 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2099 instant_txn = ovsdb_idl_txn_create(idl);
2100 HMAP_FOR_EACH (br, node, &all_bridges) {
2101 struct iface *iface;
2104 br_refresh_stp_status(br);
2106 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2107 port_refresh_stp_status(port);
2110 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2111 enum netdev_flags flags;
2112 const char *link_state;
2113 int64_t link_resets;
2116 if (iface_is_synthetic(iface)) {
2120 current = ofproto_port_is_lacp_current(br->ofproto,
2124 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2126 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2129 error = netdev_get_flags(iface->netdev, &flags);
2131 const char *state = flags & NETDEV_UP ? "up" : "down";
2132 ovsrec_interface_set_admin_state(iface->cfg, state);
2134 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2137 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2138 ovsrec_interface_set_link_state(iface->cfg, link_state);
2140 link_resets = netdev_get_carrier_resets(iface->netdev);
2141 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2143 iface_refresh_cfm_stats(iface);
2148 status = ovsdb_idl_txn_commit(instant_txn);
2149 if (status != TXN_INCOMPLETE) {
2150 ovsdb_idl_txn_destroy(instant_txn);
2153 if (status == TXN_UNCHANGED) {
2154 instant_stats_could_have_changed = false;
2159 instant_stats_wait(void)
2162 ovsdb_idl_txn_wait(instant_txn);
2163 } else if (instant_stats_could_have_changed) {
2164 poll_timer_wait_until(instant_next_txn);
2168 /* Performs periodic activity required by bridges that needs to be done with
2169 * the least possible latency.
2171 * It makes sense to call this function a couple of times per poll loop, to
2172 * provide a significant performance boost on some benchmarks with ofprotos
2173 * that use the ofproto-dpif implementation. */
2175 bridge_run_fast(void)
2182 ofproto_enumerate_types(&types);
2183 SSET_FOR_EACH (type, &types) {
2184 ofproto_type_run_fast(type);
2186 sset_destroy(&types);
2188 HMAP_FOR_EACH (br, node, &all_bridges) {
2189 ofproto_run_fast(br->ofproto);
2196 static struct ovsrec_open_vswitch null_cfg;
2197 const struct ovsrec_open_vswitch *cfg;
2198 struct ovsdb_idl_txn *reconf_txn = NULL;
2202 bool vlan_splinters_changed;
2205 ovsrec_open_vswitch_init(&null_cfg);
2207 /* (Re)configure if necessary. */
2208 if (!reconfiguring) {
2211 if (ovsdb_idl_is_lock_contended(idl)) {
2212 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2213 struct bridge *br, *next_br;
2215 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2216 "disabling this process until it goes away");
2218 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2222 } else if (!ovsdb_idl_has_lock(idl)) {
2226 cfg = ovsrec_open_vswitch_first(idl);
2228 /* Initialize the ofproto library. This only needs to run once, but
2229 * it must be done after the configuration is set. If the
2230 * initialization has already occurred, bridge_init_ofproto()
2231 * returns immediately. */
2232 bridge_init_ofproto(cfg);
2234 /* Let each datapath type do the work that it needs to do. */
2236 ofproto_enumerate_types(&types);
2237 SSET_FOR_EACH (type, &types) {
2238 ofproto_type_run(type);
2240 sset_destroy(&types);
2242 /* Let each bridge do the work that it needs to do. */
2243 HMAP_FOR_EACH (br, node, &all_bridges) {
2244 ofproto_run(br->ofproto);
2247 /* Re-configure SSL. We do this on every trip through the main loop,
2248 * instead of just when the database changes, because the contents of the
2249 * key and certificate files can change without the database changing.
2251 * We do this before bridge_reconfigure() because that function might
2252 * initiate SSL connections and thus requires SSL to be configured. */
2253 if (cfg && cfg->ssl) {
2254 const struct ovsrec_ssl *ssl = cfg->ssl;
2256 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2257 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2260 if (!reconfiguring) {
2261 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2262 * usage has changed. */
2263 vlan_splinters_changed = false;
2264 if (vlan_splinters_enabled_anywhere) {
2265 HMAP_FOR_EACH (br, node, &all_bridges) {
2266 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2267 vlan_splinters_changed = true;
2273 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2274 idl_seqno = ovsdb_idl_get_seqno(idl);
2276 reconf_txn = ovsdb_idl_txn_create(idl);
2277 bridge_reconfigure(cfg);
2279 /* We still need to reconfigure to avoid dangling pointers to
2280 * now-destroyed ovsrec structures inside bridge data. */
2281 bridge_reconfigure(&null_cfg);
2286 if (reconfiguring) {
2288 reconf_txn = ovsdb_idl_txn_create(idl);
2291 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2292 reconfiguring = false;
2295 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2298 /* If we are completing our initial configuration for this run
2299 * of ovs-vswitchd, then keep the transaction around to monitor
2300 * it for completion. */
2301 if (!initial_config_done) {
2302 initial_config_done = true;
2303 daemonize_txn = reconf_txn;
2310 ovsdb_idl_txn_commit(reconf_txn);
2311 ovsdb_idl_txn_destroy(reconf_txn);
2315 if (daemonize_txn) {
2316 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2317 if (status != TXN_INCOMPLETE) {
2318 ovsdb_idl_txn_destroy(daemonize_txn);
2319 daemonize_txn = NULL;
2321 /* ovs-vswitchd has completed initialization, so allow the
2322 * process that forked us to exit successfully. */
2323 daemonize_complete();
2325 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2329 /* Refresh interface and mirror stats if necessary. */
2330 if (time_msec() >= iface_stats_timer) {
2332 struct ovsdb_idl_txn *txn;
2334 txn = ovsdb_idl_txn_create(idl);
2335 HMAP_FOR_EACH (br, node, &all_bridges) {
2339 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2340 struct iface *iface;
2342 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2343 iface_refresh_stats(iface);
2344 iface_refresh_status(iface);
2348 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2349 mirror_refresh_stats(m);
2353 refresh_controller_status();
2354 ovsdb_idl_txn_commit(txn);
2355 ovsdb_idl_txn_destroy(txn); /* XXX */
2358 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2362 instant_stats_run();
2371 ovsdb_idl_wait(idl);
2372 if (daemonize_txn) {
2373 ovsdb_idl_txn_wait(daemonize_txn);
2376 if (reconfiguring) {
2377 poll_immediate_wake();
2381 ofproto_enumerate_types(&types);
2382 SSET_FOR_EACH (type, &types) {
2383 ofproto_type_wait(type);
2385 sset_destroy(&types);
2387 if (!hmap_is_empty(&all_bridges)) {
2390 HMAP_FOR_EACH (br, node, &all_bridges) {
2391 ofproto_wait(br->ofproto);
2393 poll_timer_wait_until(iface_stats_timer);
2396 system_stats_wait();
2397 instant_stats_wait();
2400 /* Adds some memory usage statistics for bridges into 'usage', for use with
2401 * memory_report(). */
2403 bridge_get_memory_usage(struct simap *usage)
2407 HMAP_FOR_EACH (br, node, &all_bridges) {
2408 ofproto_get_memory_usage(br->ofproto, usage);
2412 /* QoS unixctl user interface functions. */
2414 struct qos_unixctl_show_cbdata {
2416 struct iface *iface;
2420 qos_unixctl_show_cb(unsigned int queue_id,
2421 const struct smap *details,
2424 struct qos_unixctl_show_cbdata *data = aux;
2425 struct ds *ds = data->ds;
2426 struct iface *iface = data->iface;
2427 struct netdev_queue_stats stats;
2428 struct smap_node *node;
2431 ds_put_cstr(ds, "\n");
2433 ds_put_format(ds, "Queue %u:\n", queue_id);
2435 ds_put_cstr(ds, "Default:\n");
2438 SMAP_FOR_EACH (node, details) {
2439 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2442 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2444 if (stats.tx_packets != UINT64_MAX) {
2445 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2448 if (stats.tx_bytes != UINT64_MAX) {
2449 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2452 if (stats.tx_errors != UINT64_MAX) {
2453 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2456 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2457 queue_id, strerror(error));
2462 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2463 const char *argv[], void *aux OVS_UNUSED)
2465 struct ds ds = DS_EMPTY_INITIALIZER;
2466 struct smap smap = SMAP_INITIALIZER(&smap);
2467 struct iface *iface;
2469 struct smap_node *node;
2470 struct qos_unixctl_show_cbdata data;
2473 iface = iface_find(argv[1]);
2475 unixctl_command_reply_error(conn, "no such interface");
2479 netdev_get_qos(iface->netdev, &type, &smap);
2481 if (*type != '\0') {
2482 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2484 SMAP_FOR_EACH (node, &smap) {
2485 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2490 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2493 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2495 unixctl_command_reply(conn, ds_cstr(&ds));
2497 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2498 unixctl_command_reply_error(conn, ds_cstr(&ds));
2501 smap_destroy(&smap);
2505 /* Bridge reconfiguration functions. */
2507 bridge_create(const struct ovsrec_bridge *br_cfg)
2511 ovs_assert(!bridge_lookup(br_cfg->name));
2512 br = xzalloc(sizeof *br);
2514 br->name = xstrdup(br_cfg->name);
2515 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2518 /* Derive the default Ethernet address from the bridge's UUID. This should
2519 * be unique and it will be stable between ovs-vswitchd runs. */
2520 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2521 eth_addr_mark_random(br->default_ea);
2523 hmap_init(&br->ports);
2524 hmap_init(&br->ifaces);
2525 hmap_init(&br->iface_by_name);
2526 hmap_init(&br->mirrors);
2528 hmap_init(&br->if_cfg_todo);
2529 list_init(&br->ofpp_garbage);
2531 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2535 bridge_destroy(struct bridge *br)
2538 struct mirror *mirror, *next_mirror;
2539 struct port *port, *next_port;
2540 struct if_cfg *if_cfg, *next_if_cfg;
2541 struct ofpp_garbage *garbage, *next_garbage;
2543 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2546 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2547 mirror_destroy(mirror);
2549 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2550 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2553 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2554 &br->ofpp_garbage) {
2555 list_remove(&garbage->list_node);
2559 hmap_remove(&all_bridges, &br->node);
2560 ofproto_destroy(br->ofproto);
2561 hmap_destroy(&br->ifaces);
2562 hmap_destroy(&br->ports);
2563 hmap_destroy(&br->iface_by_name);
2564 hmap_destroy(&br->mirrors);
2565 hmap_destroy(&br->if_cfg_todo);
2572 static struct bridge *
2573 bridge_lookup(const char *name)
2577 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2578 if (!strcmp(br->name, name)) {
2585 /* Handle requests for a listing of all flows known by the OpenFlow
2586 * stack, including those normally hidden. */
2588 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2589 const char *argv[], void *aux OVS_UNUSED)
2594 br = bridge_lookup(argv[1]);
2596 unixctl_command_reply_error(conn, "Unknown bridge");
2601 ofproto_get_all_flows(br->ofproto, &results);
2603 unixctl_command_reply(conn, ds_cstr(&results));
2604 ds_destroy(&results);
2607 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2608 * connections and reconnect. If BRIDGE is not specified, then all bridges
2609 * drop their controller connections and reconnect. */
2611 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2612 const char *argv[], void *aux OVS_UNUSED)
2616 br = bridge_lookup(argv[1]);
2618 unixctl_command_reply_error(conn, "Unknown bridge");
2621 ofproto_reconnect_controllers(br->ofproto);
2623 HMAP_FOR_EACH (br, node, &all_bridges) {
2624 ofproto_reconnect_controllers(br->ofproto);
2627 unixctl_command_reply(conn, NULL);
2631 bridge_get_controllers(const struct bridge *br,
2632 struct ovsrec_controller ***controllersp)
2634 struct ovsrec_controller **controllers;
2635 size_t n_controllers;
2637 controllers = br->cfg->controller;
2638 n_controllers = br->cfg->n_controller;
2640 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2646 *controllersp = controllers;
2648 return n_controllers;
2652 bridge_queue_if_cfg(struct bridge *br,
2653 const struct ovsrec_interface *cfg,
2654 const struct ovsrec_port *parent)
2656 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2659 if_cfg->parent = parent;
2660 if_cfg->ofport = iface_pick_ofport(cfg);
2661 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2662 hash_string(if_cfg->cfg->name, 0));
2665 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2666 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2667 * 'br' needs to complete its configuration. */
2669 bridge_add_del_ports(struct bridge *br,
2670 const unsigned long int *splinter_vlans)
2672 struct shash_node *port_node;
2673 struct port *port, *next;
2674 struct shash new_ports;
2677 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2679 /* Collect new ports. */
2680 shash_init(&new_ports);
2681 for (i = 0; i < br->cfg->n_ports; i++) {
2682 const char *name = br->cfg->ports[i]->name;
2683 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2684 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2688 if (bridge_get_controllers(br, NULL)
2689 && !shash_find(&new_ports, br->name)) {
2690 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2691 br->name, br->name);
2693 ovsrec_interface_init(&br->synth_local_iface);
2694 ovsrec_port_init(&br->synth_local_port);
2696 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2697 br->synth_local_port.n_interfaces = 1;
2698 br->synth_local_port.name = br->name;
2700 br->synth_local_iface.name = br->name;
2701 br->synth_local_iface.type = "internal";
2703 br->synth_local_ifacep = &br->synth_local_iface;
2705 shash_add(&new_ports, br->name, &br->synth_local_port);
2708 if (splinter_vlans) {
2709 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2712 /* Get rid of deleted ports.
2713 * Get rid of deleted interfaces on ports that still exist. */
2714 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2715 port->cfg = shash_find_data(&new_ports, port->name);
2719 port_del_ifaces(port);
2723 /* Update iface->cfg and iface->type in interfaces that still exist.
2724 * Add new interfaces to creation queue. */
2725 SHASH_FOR_EACH (port_node, &new_ports) {
2726 const struct ovsrec_port *port = port_node->data;
2729 for (i = 0; i < port->n_interfaces; i++) {
2730 const struct ovsrec_interface *cfg = port->interfaces[i];
2731 struct iface *iface = iface_lookup(br, cfg->name);
2732 const char *type = iface_get_type(cfg, br->cfg);
2737 } else if (!strcmp(type, "null")) {
2738 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2739 " may be removed in February 2013. Please email"
2740 " dev@openvswitch.org with concerns.",
2743 bridge_queue_if_cfg(br, cfg, port);
2748 shash_destroy(&new_ports);
2751 /* Initializes 'oc' appropriately as a management service controller for
2754 * The caller must free oc->target when it is no longer needed. */
2756 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2757 struct ofproto_controller *oc)
2759 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2760 oc->max_backoff = 0;
2761 oc->probe_interval = 60;
2762 oc->band = OFPROTO_OUT_OF_BAND;
2764 oc->burst_limit = 0;
2765 oc->enable_async_msgs = true;
2768 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2770 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2771 struct ofproto_controller *oc)
2775 oc->target = c->target;
2776 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2777 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2778 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2779 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2780 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2781 oc->burst_limit = (c->controller_burst_limit
2782 ? *c->controller_burst_limit : 0);
2783 oc->enable_async_msgs = (!c->enable_async_messages
2784 || *c->enable_async_messages);
2785 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2786 if (dscp < 0 || dscp > 63) {
2787 dscp = DSCP_DEFAULT;
2792 /* Configures the IP stack for 'br''s local interface properly according to the
2793 * configuration in 'c'. */
2795 bridge_configure_local_iface_netdev(struct bridge *br,
2796 struct ovsrec_controller *c)
2798 struct netdev *netdev;
2799 struct in_addr mask, gateway;
2801 struct iface *local_iface;
2804 /* If there's no local interface or no IP address, give up. */
2805 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2806 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2810 /* Bring up the local interface. */
2811 netdev = local_iface->netdev;
2812 netdev_turn_flags_on(netdev, NETDEV_UP, true);
2814 /* Configure the IP address and netmask. */
2815 if (!c->local_netmask
2816 || !inet_aton(c->local_netmask, &mask)
2818 mask.s_addr = guess_netmask(ip.s_addr);
2820 if (!netdev_set_in4(netdev, ip, mask)) {
2821 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2822 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2825 /* Configure the default gateway. */
2826 if (c->local_gateway
2827 && inet_aton(c->local_gateway, &gateway)
2828 && gateway.s_addr) {
2829 if (!netdev_add_router(netdev, gateway)) {
2830 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2831 br->name, IP_ARGS(gateway.s_addr));
2836 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2837 * in either string are treated as equal to any number of slashes in the other,
2838 * e.g. "x///y" is equal to "x/y".
2840 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2841 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2842 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2843 * 'b' against a prefix of 'a'.
2846 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2848 const char *b_start = b;
2850 if (b - b_start >= b_stoplen) {
2852 } else if (*a != *b) {
2854 } else if (*a == '/') {
2855 a += strspn(a, "/");
2856 b += strspn(b, "/");
2857 } else if (*a == '\0') {
2867 bridge_configure_remotes(struct bridge *br,
2868 const struct sockaddr_in *managers, size_t n_managers)
2870 bool disable_in_band;
2872 struct ovsrec_controller **controllers;
2873 size_t n_controllers;
2875 enum ofproto_fail_mode fail_mode;
2877 struct ofproto_controller *ocs;
2881 /* Check if we should disable in-band control on this bridge. */
2882 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2885 /* Set OpenFlow queue ID for in-band control. */
2886 ofproto_set_in_band_queue(br->ofproto,
2887 smap_get_int(&br->cfg->other_config,
2888 "in-band-queue", -1));
2890 if (disable_in_band) {
2891 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2893 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2896 n_controllers = bridge_get_controllers(br, &controllers);
2898 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2901 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2902 for (i = 0; i < n_controllers; i++) {
2903 struct ovsrec_controller *c = controllers[i];
2905 if (!strncmp(c->target, "punix:", 6)
2906 || !strncmp(c->target, "unix:", 5)) {
2907 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2910 if (!strncmp(c->target, "unix:", 5)) {
2911 /* Connect to a listening socket */
2912 whitelist = xasprintf("unix:%s/", ovs_rundir());
2913 if (strchr(c->target, '/') &&
2914 !equal_pathnames(c->target, whitelist,
2915 strlen(whitelist))) {
2916 /* Absolute path specified, but not in ovs_rundir */
2917 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2918 "controller \"%s\" due to possibility for "
2919 "remote exploit. Instead, specify socket "
2920 "in whitelisted \"%s\" or connect to "
2921 "\"unix:%s/%s.mgmt\" (which is always "
2922 "available without special configuration).",
2923 br->name, c->target, whitelist,
2924 ovs_rundir(), br->name);
2929 whitelist = xasprintf("punix:%s/%s.controller",
2930 ovs_rundir(), br->name);
2931 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
2932 /* Prevent remote ovsdb-server users from accessing
2933 * arbitrary Unix domain sockets and overwriting arbitrary
2935 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
2936 "controller \"%s\" due to possibility of "
2937 "overwriting local files. Instead, specify "
2938 "whitelisted \"%s\" or connect to "
2939 "\"unix:%s/%s.mgmt\" (which is always "
2940 "available without special configuration).",
2941 br->name, c->target, whitelist,
2942 ovs_rundir(), br->name);
2951 bridge_configure_local_iface_netdev(br, c);
2952 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
2953 if (disable_in_band) {
2954 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
2959 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
2960 bridge_get_allowed_versions(br));
2961 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
2964 /* Set the fail-mode. */
2965 fail_mode = !br->cfg->fail_mode
2966 || !strcmp(br->cfg->fail_mode, "standalone")
2967 ? OFPROTO_FAIL_STANDALONE
2968 : OFPROTO_FAIL_SECURE;
2969 ofproto_set_fail_mode(br->ofproto, fail_mode);
2971 /* Configure OpenFlow controller connection snooping. */
2972 if (!ofproto_has_snoops(br->ofproto)) {
2976 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
2977 ovs_rundir(), br->name));
2978 ofproto_set_snoops(br->ofproto, &snoops);
2979 sset_destroy(&snoops);
2984 bridge_configure_tables(struct bridge *br)
2986 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2990 n_tables = ofproto_get_n_tables(br->ofproto);
2992 for (i = 0; i < n_tables; i++) {
2993 struct ofproto_table_settings s;
2996 s.max_flows = UINT_MAX;
3000 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3001 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3004 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3005 s.max_flows = *cfg->flow_limit;
3007 if (cfg->overflow_policy
3008 && !strcmp(cfg->overflow_policy, "evict")) {
3011 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3012 for (k = 0; k < cfg->n_groups; k++) {
3013 const char *string = cfg->groups[k];
3016 msg = mf_parse_subfield__(&s.groups[k], &string);
3018 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3019 "'groups' (%s)", br->name, i, msg);
3021 } else if (*string) {
3022 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3023 "element '%s' contains trailing garbage",
3024 br->name, i, cfg->groups[k]);
3032 ofproto_configure_table(br->ofproto, i, &s);
3036 for (; j < br->cfg->n_flow_tables; j++) {
3037 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3038 "%"PRId64" not supported by this datapath", br->name,
3039 br->cfg->key_flow_tables[j]);
3044 bridge_configure_dp_desc(struct bridge *br)
3046 ofproto_set_dp_desc(br->ofproto,
3047 smap_get(&br->cfg->other_config, "dp-desc"));
3050 /* Port functions. */
3052 static struct port *
3053 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3057 port = xzalloc(sizeof *port);
3059 port->name = xstrdup(cfg->name);
3061 list_init(&port->ifaces);
3063 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3067 /* Deletes interfaces from 'port' that are no longer configured for it. */
3069 port_del_ifaces(struct port *port)
3071 struct iface *iface, *next;
3072 struct sset new_ifaces;
3075 /* Collect list of new interfaces. */
3076 sset_init(&new_ifaces);
3077 for (i = 0; i < port->cfg->n_interfaces; i++) {
3078 const char *name = port->cfg->interfaces[i]->name;
3079 const char *type = port->cfg->interfaces[i]->type;
3080 if (strcmp(type, "null")) {
3081 sset_add(&new_ifaces, name);
3085 /* Get rid of deleted interfaces. */
3086 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3087 if (!sset_contains(&new_ifaces, iface->name)) {
3088 iface_destroy(iface);
3092 sset_destroy(&new_ifaces);
3096 port_destroy(struct port *port)
3099 struct bridge *br = port->bridge;
3100 struct iface *iface, *next;
3103 ofproto_bundle_unregister(br->ofproto, port);
3106 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3107 iface_destroy(iface);
3110 hmap_remove(&br->ports, &port->hmap_node);
3116 static struct port *
3117 port_lookup(const struct bridge *br, const char *name)
3121 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3123 if (!strcmp(port->name, name)) {
3131 enable_lacp(struct port *port, bool *activep)
3133 if (!port->cfg->lacp) {
3134 /* XXX when LACP implementation has been sufficiently tested, enable by
3135 * default and make active on bonded ports. */
3137 } else if (!strcmp(port->cfg->lacp, "off")) {
3139 } else if (!strcmp(port->cfg->lacp, "active")) {
3142 } else if (!strcmp(port->cfg->lacp, "passive")) {
3146 VLOG_WARN("port %s: unknown LACP mode %s",
3147 port->name, port->cfg->lacp);
3152 static struct lacp_settings *
3153 port_configure_lacp(struct port *port, struct lacp_settings *s)
3155 const char *lacp_time, *system_id;
3158 if (!enable_lacp(port, &s->active)) {
3162 s->name = port->name;
3164 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3166 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3167 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3168 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3169 " address.", port->name, system_id);
3173 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3176 if (eth_addr_is_zero(s->id)) {
3177 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3181 /* Prefer bondable links if unspecified. */
3182 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3184 s->priority = (priority > 0 && priority <= UINT16_MAX
3186 : UINT16_MAX - !list_is_short(&port->ifaces));
3188 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3189 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3194 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3196 int priority, portid, key;
3198 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3199 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3201 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3203 if (portid <= 0 || portid > UINT16_MAX) {
3204 portid = iface->ofp_port;
3207 if (priority <= 0 || priority > UINT16_MAX) {
3208 priority = UINT16_MAX;
3211 if (key < 0 || key > UINT16_MAX) {
3215 s->name = iface->name;
3217 s->priority = priority;
3222 port_configure_bond(struct port *port, struct bond_settings *s,
3223 uint32_t *bond_stable_ids)
3225 const char *detect_s;
3226 struct iface *iface;
3227 int miimon_interval;
3230 s->name = port->name;
3232 if (port->cfg->bond_mode) {
3233 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3234 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3235 port->name, port->cfg->bond_mode,
3236 bond_mode_to_string(s->balance));
3239 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3241 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3242 * active-backup. At some point we should remove this warning. */
3243 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3244 " in previous versions, the default bond_mode was"
3245 " balance-slb", port->name,
3246 bond_mode_to_string(s->balance));
3248 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3249 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3250 "please use another bond type or disable flood_vlans",
3254 miimon_interval = smap_get_int(&port->cfg->other_config,
3255 "bond-miimon-interval", 0);
3256 if (miimon_interval <= 0) {
3257 miimon_interval = 200;
3260 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3261 if (!detect_s || !strcmp(detect_s, "carrier")) {
3262 miimon_interval = 0;
3263 } else if (strcmp(detect_s, "miimon")) {
3264 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3265 "defaulting to carrier", port->name, detect_s);
3266 miimon_interval = 0;
3269 s->up_delay = MAX(0, port->cfg->bond_updelay);
3270 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3271 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3272 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3273 "bond-rebalance-interval", 10000);
3274 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3275 s->rebalance_interval = 1000;
3278 s->fake_iface = port->cfg->bond_fake_iface;
3281 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3282 long long stable_id;
3284 stable_id = smap_get_int(&iface->cfg->other_config, "bond-stable-id",
3286 if (stable_id <= 0 || stable_id >= UINT32_MAX) {
3287 stable_id = iface->ofp_port;
3289 bond_stable_ids[i++] = stable_id;
3291 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3295 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3296 * instead of obtaining it from the database. */
3298 port_is_synthetic(const struct port *port)
3300 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3303 /* Interface functions. */
3306 iface_is_internal(const struct ovsrec_interface *iface,
3307 const struct ovsrec_bridge *br)
3309 /* The local port and "internal" ports are always "internal". */
3310 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3313 /* Returns the correct network device type for interface 'iface' in bridge
3316 iface_get_type(const struct ovsrec_interface *iface,
3317 const struct ovsrec_bridge *br)
3321 /* The local port always has type "internal". Other ports take
3322 * their type from the database and default to "system" if none is
3324 if (iface_is_internal(iface, br)) {
3327 type = iface->type[0] ? iface->type : "system";
3330 return ofproto_port_open_type(br->datapath_type, type);
3334 iface_destroy(struct iface *iface)
3337 struct port *port = iface->port;
3338 struct bridge *br = port->bridge;
3340 if (br->ofproto && iface->ofp_port >= 0) {
3341 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3344 if (iface->ofp_port >= 0) {
3345 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3348 list_remove(&iface->port_elem);
3349 hmap_remove(&br->iface_by_name, &iface->name_node);
3351 netdev_close(iface->netdev);
3358 static struct iface *
3359 iface_lookup(const struct bridge *br, const char *name)
3361 struct iface *iface;
3363 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3364 &br->iface_by_name) {
3365 if (!strcmp(iface->name, name)) {
3373 static struct iface *
3374 iface_find(const char *name)
3376 const struct bridge *br;
3378 HMAP_FOR_EACH (br, node, &all_bridges) {
3379 struct iface *iface = iface_lookup(br, name);
3388 static struct if_cfg *
3389 if_cfg_lookup(const struct bridge *br, const char *name)
3391 struct if_cfg *if_cfg;
3393 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3395 if (!strcmp(if_cfg->cfg->name, name)) {
3403 static struct iface *
3404 iface_from_ofp_port(const struct bridge *br, uint16_t ofp_port)
3406 struct iface *iface;
3408 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node,
3409 hash_int(ofp_port, 0), &br->ifaces) {
3410 if (iface->ofp_port == ofp_port) {
3417 /* Set Ethernet address of 'iface', if one is specified in the configuration
3420 iface_set_mac(struct iface *iface)
3422 uint8_t ea[ETH_ADDR_LEN];
3424 if (!strcmp(iface->type, "internal")
3425 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3426 if (iface->ofp_port == OFPP_LOCAL) {
3427 VLOG_ERR("interface %s: ignoring mac in Interface record "
3428 "(use Bridge record to set local port's mac)",
3430 } else if (eth_addr_is_multicast(ea)) {
3431 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3434 int error = netdev_set_etheraddr(iface->netdev, ea);
3436 VLOG_ERR("interface %s: setting MAC failed (%s)",
3437 iface->name, strerror(error));
3443 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3445 iface_set_ofport(const struct ovsrec_interface *if_cfg, int64_t ofport)
3447 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3448 ovsrec_interface_set_ofport(if_cfg, &ofport, 1);
3452 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3453 * sets the "ofport" field to -1.
3455 * This is appropriate when 'if_cfg''s interface cannot be created or is
3456 * otherwise invalid. */
3458 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3460 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3461 ovsrec_interface_set_status(if_cfg, NULL);
3462 ovsrec_interface_set_admin_state(if_cfg, NULL);
3463 ovsrec_interface_set_duplex(if_cfg, NULL);
3464 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3465 ovsrec_interface_set_link_state(if_cfg, NULL);
3466 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3467 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3468 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3469 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3470 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3471 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3472 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3476 struct iface_delete_queues_cbdata {
3477 struct netdev *netdev;
3478 const struct ovsdb_datum *queues;
3482 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3484 union ovsdb_atom atom;
3486 atom.integer = target;
3487 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3491 iface_delete_queues(unsigned int queue_id,
3492 const struct smap *details OVS_UNUSED, void *cbdata_)
3494 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3496 if (!queue_ids_include(cbdata->queues, queue_id)) {
3497 netdev_delete_queue(cbdata->netdev, queue_id);
3502 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3504 struct ofpbuf queues_buf;
3506 ofpbuf_init(&queues_buf, 0);
3508 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3509 netdev_set_qos(iface->netdev, NULL, NULL);
3511 struct iface_delete_queues_cbdata cbdata;
3515 /* Configure top-level Qos for 'iface'. */
3516 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3518 /* Deconfigure queues that were deleted. */
3519 cbdata.netdev = iface->netdev;
3520 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3522 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3524 /* Configure queues for 'iface'. */
3526 for (i = 0; i < qos->n_queues; i++) {
3527 const struct ovsrec_queue *queue = qos->value_queues[i];
3528 unsigned int queue_id = qos->key_queues[i];
3530 if (queue_id == 0) {
3534 if (queue->n_dscp == 1) {
3535 struct ofproto_port_queue *port_queue;
3537 port_queue = ofpbuf_put_uninit(&queues_buf,
3538 sizeof *port_queue);
3539 port_queue->queue = queue_id;
3540 port_queue->dscp = queue->dscp[0];
3543 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3546 struct smap details;
3548 smap_init(&details);
3549 netdev_set_queue(iface->netdev, 0, &details);
3550 smap_destroy(&details);
3554 if (iface->ofp_port >= 0) {
3555 const struct ofproto_port_queue *port_queues = queues_buf.data;
3556 size_t n_queues = queues_buf.size / sizeof *port_queues;
3558 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3559 port_queues, n_queues);
3562 netdev_set_policing(iface->netdev,
3563 iface->cfg->ingress_policing_rate,
3564 iface->cfg->ingress_policing_burst);
3566 ofpbuf_uninit(&queues_buf);
3570 iface_configure_cfm(struct iface *iface)
3572 const struct ovsrec_interface *cfg = iface->cfg;
3573 const char *opstate_str;
3574 const char *cfm_ccm_vlan;
3575 struct cfm_settings s;
3576 struct smap netdev_args;
3578 if (!cfg->n_cfm_mpid) {
3579 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3583 s.check_tnl_key = false;
3584 smap_init(&netdev_args);
3585 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3586 const char *key = smap_get(&netdev_args, "key");
3587 const char *in_key = smap_get(&netdev_args, "in_key");
3589 s.check_tnl_key = (key && !strcmp(key, "flow"))
3590 || (in_key && !strcmp(in_key, "flow"));
3592 smap_destroy(&netdev_args);
3594 s.mpid = *cfg->cfm_mpid;
3595 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3596 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3597 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3599 if (s.interval <= 0) {
3603 if (!cfm_ccm_vlan) {
3605 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3606 s.ccm_vlan = CFM_RANDOM_VLAN;
3608 s.ccm_vlan = atoi(cfm_ccm_vlan);
3609 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3614 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3617 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3618 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3620 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3623 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3624 * instead of obtaining it from the database. */
3626 iface_is_synthetic(const struct iface *iface)
3628 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3632 iface_pick_ofport(const struct ovsrec_interface *cfg)
3634 int64_t ofport = cfg->n_ofport ? *cfg->ofport : OFPP_NONE;
3635 return cfg->n_ofport_request ? *cfg->ofport_request : ofport;
3639 /* Port mirroring. */
3641 static struct mirror *
3642 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3646 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3647 if (uuid_equals(uuid, &m->uuid)) {
3655 bridge_configure_mirrors(struct bridge *br)
3657 const struct ovsdb_datum *mc;
3658 unsigned long *flood_vlans;
3659 struct mirror *m, *next;
3662 /* Get rid of deleted mirrors. */
3663 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3664 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3665 union ovsdb_atom atom;
3667 atom.uuid = m->uuid;
3668 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3673 /* Add new mirrors and reconfigure existing ones. */
3674 for (i = 0; i < br->cfg->n_mirrors; i++) {
3675 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3676 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3678 m = mirror_create(br, cfg);
3681 if (!mirror_configure(m)) {
3686 /* Update flooded vlans (for RSPAN). */
3687 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3688 br->cfg->n_flood_vlans);
3689 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3690 bitmap_free(flood_vlans);
3693 static struct mirror *
3694 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3698 m = xzalloc(sizeof *m);
3699 m->uuid = cfg->header_.uuid;
3700 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3702 m->name = xstrdup(cfg->name);
3708 mirror_destroy(struct mirror *m)
3711 struct bridge *br = m->bridge;
3714 ofproto_mirror_unregister(br->ofproto, m);
3717 hmap_remove(&br->mirrors, &m->hmap_node);
3724 mirror_collect_ports(struct mirror *m,
3725 struct ovsrec_port **in_ports, int n_in_ports,
3726 void ***out_portsp, size_t *n_out_portsp)
3728 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3729 size_t n_out_ports = 0;
3732 for (i = 0; i < n_in_ports; i++) {
3733 const char *name = in_ports[i]->name;
3734 struct port *port = port_lookup(m->bridge, name);
3736 out_ports[n_out_ports++] = port;
3738 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3739 "port %s", m->bridge->name, m->name, name);
3742 *out_portsp = out_ports;
3743 *n_out_portsp = n_out_ports;
3747 mirror_configure(struct mirror *m)
3749 const struct ovsrec_mirror *cfg = m->cfg;
3750 struct ofproto_mirror_settings s;
3753 if (strcmp(cfg->name, m->name)) {
3755 m->name = xstrdup(cfg->name);
3759 /* Get output port or VLAN. */
3760 if (cfg->output_port) {
3761 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3762 if (!s.out_bundle) {
3763 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3764 m->bridge->name, m->name);
3767 s.out_vlan = UINT16_MAX;
3769 if (cfg->output_vlan) {
3770 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3771 "output vlan; ignoring output vlan",
3772 m->bridge->name, m->name);
3774 } else if (cfg->output_vlan) {
3775 /* The database should prevent invalid VLAN values. */
3776 s.out_bundle = NULL;
3777 s.out_vlan = *cfg->output_vlan;
3779 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3780 m->bridge->name, m->name);
3784 /* Get port selection. */
3785 if (cfg->select_all) {
3786 size_t n_ports = hmap_count(&m->bridge->ports);
3787 void **ports = xmalloc(n_ports * sizeof *ports);
3792 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3802 /* Get ports, dropping ports that don't exist.
3803 * The IDL ensures that there are no duplicates. */
3804 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3805 &s.srcs, &s.n_srcs);
3806 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3807 &s.dsts, &s.n_dsts);
3810 /* Get VLAN selection. */
3811 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3814 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3817 if (s.srcs != s.dsts) {
3826 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3828 * This is deprecated. It is only for compatibility with broken device drivers
3829 * in old versions of Linux that do not properly support VLANs when VLAN
3830 * devices are not used. When broken device drivers are no longer in
3831 * widespread use, we will delete these interfaces. */
3833 static struct ovsrec_port **recs;
3834 static size_t n_recs, allocated_recs;
3836 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3837 * splinters are reconfigured. */
3839 register_rec(struct ovsrec_port *rec)
3841 if (n_recs >= allocated_recs) {
3842 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3844 recs[n_recs++] = rec;
3847 /* Frees all of the ports registered with register_reg(). */
3849 free_registered_recs(void)
3853 for (i = 0; i < n_recs; i++) {
3854 struct ovsrec_port *port = recs[i];
3857 for (j = 0; j < port->n_interfaces; j++) {
3858 struct ovsrec_interface *iface = port->interfaces[j];
3863 smap_destroy(&port->other_config);
3864 free(port->interfaces);
3872 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3875 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3877 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3881 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3884 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3885 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3886 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3889 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3890 * use, returns NULL.
3892 * Updates 'vlan_splinters_enabled_anywhere'. */
3893 static unsigned long int *
3894 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
3896 unsigned long int *splinter_vlans;
3897 struct sset splinter_ifaces;
3898 const char *real_dev_name;
3899 struct shash *real_devs;
3900 struct shash_node *node;
3904 /* Free space allocated for synthesized ports and interfaces, since we're
3905 * in the process of reconstructing all of them. */
3906 free_registered_recs();
3908 splinter_vlans = bitmap_allocate(4096);
3909 sset_init(&splinter_ifaces);
3910 vlan_splinters_enabled_anywhere = false;
3911 for (i = 0; i < ovs_cfg->n_bridges; i++) {
3912 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
3915 for (j = 0; j < br_cfg->n_ports; j++) {
3916 struct ovsrec_port *port_cfg = br_cfg->ports[j];
3919 for (k = 0; k < port_cfg->n_interfaces; k++) {
3920 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
3922 if (vlan_splinters_is_enabled(iface_cfg)) {
3923 vlan_splinters_enabled_anywhere = true;
3924 sset_add(&splinter_ifaces, iface_cfg->name);
3925 vlan_bitmap_from_array__(port_cfg->trunks,
3931 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
3932 bitmap_set1(splinter_vlans, *port_cfg->tag);
3937 if (!vlan_splinters_enabled_anywhere) {
3938 free(splinter_vlans);
3939 sset_destroy(&splinter_ifaces);
3943 HMAP_FOR_EACH (br, node, &all_bridges) {
3945 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
3949 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
3950 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
3951 * device to be created for it. */
3952 bitmap_set0(splinter_vlans, 0);
3953 bitmap_set0(splinter_vlans, 4095);
3955 /* Delete all VLAN devices that we don't need. */
3957 real_devs = vlandev_get_real_devs();
3958 SHASH_FOR_EACH (node, real_devs) {
3959 const struct vlan_real_dev *real_dev = node->data;
3960 const struct vlan_dev *vlan_dev;
3961 bool real_dev_has_splinters;
3963 real_dev_has_splinters = sset_contains(&splinter_ifaces,
3965 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
3966 if (!real_dev_has_splinters
3967 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
3968 struct netdev *netdev;
3970 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
3971 if (!netdev_get_in4(netdev, NULL, NULL) ||
3972 !netdev_get_in6(netdev, NULL)) {
3973 vlandev_del(vlan_dev->name);
3975 /* It has an IP address configured, so we don't own
3976 * it. Don't delete it. */
3978 netdev_close(netdev);
3985 /* Add all VLAN devices that we need. */
3986 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
3989 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
3990 if (!vlandev_get_name(real_dev_name, vid)) {
3991 vlandev_add(real_dev_name, vid);
3998 sset_destroy(&splinter_ifaces);
4000 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4001 free(splinter_vlans);
4004 return splinter_vlans;
4007 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4010 configure_splinter_port(struct port *port)
4012 struct ofproto *ofproto = port->bridge->ofproto;
4013 uint16_t realdev_ofp_port;
4014 const char *realdev_name;
4015 struct iface *vlandev, *realdev;
4017 ofproto_bundle_unregister(port->bridge->ofproto, port);
4019 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4022 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4023 realdev = iface_lookup(port->bridge, realdev_name);
4024 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4026 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4030 static struct ovsrec_port *
4031 synthesize_splinter_port(const char *real_dev_name,
4032 const char *vlan_dev_name, int vid)
4034 struct ovsrec_interface *iface;
4035 struct ovsrec_port *port;
4037 iface = xmalloc(sizeof *iface);
4038 ovsrec_interface_init(iface);
4039 iface->name = xstrdup(vlan_dev_name);
4040 iface->type = "system";
4042 port = xmalloc(sizeof *port);
4043 ovsrec_port_init(port);
4044 port->interfaces = xmemdup(&iface, sizeof iface);
4045 port->n_interfaces = 1;
4046 port->name = xstrdup(vlan_dev_name);
4047 port->vlan_mode = "splinter";
4048 port->tag = xmalloc(sizeof *port->tag);
4051 smap_add(&port->other_config, "realdev", real_dev_name);
4057 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4058 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4059 * 1-bit in the 'splinter_vlans' bitmap. */
4061 add_vlan_splinter_ports(struct bridge *br,
4062 const unsigned long int *splinter_vlans,
4063 struct shash *ports)
4067 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4068 * we're modifying 'ports'. */
4069 for (i = 0; i < br->cfg->n_ports; i++) {
4070 const char *name = br->cfg->ports[i]->name;
4071 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4074 for (j = 0; j < port_cfg->n_interfaces; j++) {
4075 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4077 if (vlan_splinters_is_enabled(iface_cfg)) {
4078 const char *real_dev_name;
4081 real_dev_name = iface_cfg->name;
4082 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4083 const char *vlan_dev_name;
4085 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4087 && !shash_find(ports, vlan_dev_name)) {
4088 shash_add(ports, vlan_dev_name,
4089 synthesize_splinter_port(
4090 real_dev_name, vlan_dev_name, vid));
4099 mirror_refresh_stats(struct mirror *m)
4101 struct ofproto *ofproto = m->bridge->ofproto;
4102 uint64_t tx_packets, tx_bytes;
4105 size_t stat_cnt = 0;
4107 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4108 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4112 if (tx_packets != UINT64_MAX) {
4113 keys[stat_cnt] = "tx_packets";
4114 values[stat_cnt] = tx_packets;
4117 if (tx_bytes != UINT64_MAX) {
4118 keys[stat_cnt] = "tx_bytes";
4119 values[stat_cnt] = tx_bytes;
4123 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);