1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
47 #include "socket-util.h"
49 #include "stream-ssl.h"
51 #include "system-stats.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
62 VLOG_DEFINE_THIS_MODULE(bridge);
64 COVERAGE_DEFINE(bridge_reconfigure);
66 /* Configuration of an uninstantiated iface. */
68 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
69 const struct ovsrec_interface *cfg; /* Interface record. */
70 const struct ovsrec_port *parent; /* Parent port record. */
71 ofp_port_t ofport; /* Requested OpenFlow port number. */
74 /* OpenFlow port slated for removal from ofproto. */
76 struct list list_node; /* Node in bridge's ofpp_garbage. */
77 ofp_port_t ofp_port; /* Port to be deleted. */
81 /* These members are always valid. */
82 struct list port_elem; /* Element in struct port's "ifaces" list. */
83 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
87 /* These members are valid only after bridge_reconfigure() causes them to
89 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
90 ofp_port_t ofp_port; /* OpenFlow port number, */
91 /* OFPP_NONE if unknown. */
92 struct netdev *netdev; /* Network device. */
93 const char *type; /* Usually same as cfg->type. */
94 const struct ovsrec_interface *cfg;
98 struct uuid uuid; /* UUID of this "mirror" record in database. */
99 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
100 struct bridge *bridge;
102 const struct ovsrec_mirror *cfg;
106 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
107 struct bridge *bridge;
110 const struct ovsrec_port *cfg;
112 /* An ordinary bridge port has 1 interface.
113 * A bridge port for bonding has at least 2 interfaces. */
114 struct list ifaces; /* List of "struct iface"s. */
118 struct hmap_node node; /* In 'all_bridges'. */
119 char *name; /* User-specified arbitrary name. */
120 char *type; /* Datapath type. */
121 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
122 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
123 const struct ovsrec_bridge *cfg;
125 /* OpenFlow switch processing. */
126 struct ofproto *ofproto; /* OpenFlow switch. */
129 struct hmap ports; /* "struct port"s indexed by name. */
130 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
131 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
133 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
134 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
135 Indexed on 'cfg->name'. */
137 /* Port mirroring. */
138 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
146 /* All bridges, indexed by name. */
147 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
149 /* OVSDB IDL used to obtain configuration. */
150 static struct ovsdb_idl *idl;
152 /* We want to complete daemonization, fully detaching from our parent process,
153 * only after we have completed our initial configuration, committed our state
154 * to the database, and received confirmation back from the database server
155 * that it applied the commit. This allows our parent process to know that,
156 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
157 * to have already been filled in. (It is only "very likely" rather than
158 * certain because there is always a slim possibility that the transaction will
159 * fail or that some other client has added new bridges, ports, etc. while
160 * ovs-vswitchd was configuring using an old configuration.)
162 * We only need to do this once for our initial configuration at startup, so
163 * 'initial_config_done' tracks whether we've already done it. While we are
164 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
165 * itself and is otherwise NULL. */
166 static bool initial_config_done;
167 static struct ovsdb_idl_txn *daemonize_txn;
169 /* Most recently processed IDL sequence number. */
170 static unsigned int idl_seqno;
172 /* Each time this timer expires, the bridge fetches interface and mirror
173 * statistics and pushes them into the database. */
174 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
175 static long long int iface_stats_timer = LLONG_MIN;
177 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
178 * expensive. This can cause bridge_reconfigure() to take a long time during
179 * which no other work can be done. To deal with this problem, we limit port
180 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
181 * call to bridge_reconfigure(). If there is more work to do after the limit
182 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
183 * This allows the rest of the code to catch up on important things like
184 * forwarding packets. */
185 #define OFP_PORT_ACTION_WINDOW 10
186 static bool reconfiguring = false;
188 static void add_del_bridges(const struct ovsrec_open_vswitch *);
189 static void bridge_update_ofprotos(void);
190 static void bridge_create(const struct ovsrec_bridge *);
191 static void bridge_destroy(struct bridge *);
192 static struct bridge *bridge_lookup(const char *name);
193 static unixctl_cb_func bridge_unixctl_dump_flows;
194 static unixctl_cb_func bridge_unixctl_reconnect;
195 static size_t bridge_get_controllers(const struct bridge *br,
196 struct ovsrec_controller ***controllersp);
197 static void bridge_add_del_ports(struct bridge *,
198 const unsigned long int *splinter_vlans);
199 static void bridge_refresh_ofp_port(struct bridge *);
200 static void bridge_configure_flow_miss_model(const char *opt);
201 static void bridge_configure_datapath_id(struct bridge *);
202 static void bridge_configure_netflow(struct bridge *);
203 static void bridge_configure_forward_bpdu(struct bridge *);
204 static void bridge_configure_mac_table(struct bridge *);
205 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
206 static void bridge_configure_ipfix(struct bridge *);
207 static void bridge_configure_stp(struct bridge *);
208 static void bridge_configure_tables(struct bridge *);
209 static void bridge_configure_dp_desc(struct bridge *);
210 static void bridge_configure_remotes(struct bridge *,
211 const struct sockaddr_in *managers,
213 static void bridge_pick_local_hw_addr(struct bridge *,
214 uint8_t ea[ETH_ADDR_LEN],
215 struct iface **hw_addr_iface);
216 static uint64_t bridge_pick_datapath_id(struct bridge *,
217 const uint8_t bridge_ea[ETH_ADDR_LEN],
218 struct iface *hw_addr_iface);
219 static void bridge_queue_if_cfg(struct bridge *,
220 const struct ovsrec_interface *,
221 const struct ovsrec_port *);
222 static uint64_t dpid_from_hash(const void *, size_t nbytes);
223 static bool bridge_has_bond_fake_iface(const struct bridge *,
225 static bool port_is_bond_fake_iface(const struct port *);
227 static unixctl_cb_func qos_unixctl_show;
229 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
230 static void port_del_ifaces(struct port *);
231 static void port_destroy(struct port *);
232 static struct port *port_lookup(const struct bridge *, const char *name);
233 static void port_configure(struct port *);
234 static struct lacp_settings *port_configure_lacp(struct port *,
235 struct lacp_settings *);
236 static void port_configure_bond(struct port *, struct bond_settings *);
237 static bool port_is_synthetic(const struct port *);
239 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
240 static void run_system_stats(void);
242 static void bridge_configure_mirrors(struct bridge *);
243 static struct mirror *mirror_create(struct bridge *,
244 const struct ovsrec_mirror *);
245 static void mirror_destroy(struct mirror *);
246 static bool mirror_configure(struct mirror *);
247 static void mirror_refresh_stats(struct mirror *);
249 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
250 static bool iface_create(struct bridge *, struct if_cfg *,
251 ofp_port_t ofp_port);
252 static bool iface_is_internal(const struct ovsrec_interface *iface,
253 const struct ovsrec_bridge *br);
254 static const char *iface_get_type(const struct ovsrec_interface *,
255 const struct ovsrec_bridge *);
256 static void iface_destroy(struct iface *);
257 static struct iface *iface_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_find(const char *name);
259 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
260 static struct iface *iface_from_ofp_port(const struct bridge *,
261 ofp_port_t ofp_port);
262 static void iface_set_mac(struct iface *);
263 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
264 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
265 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
266 static void iface_configure_cfm(struct iface *);
267 static void iface_refresh_cfm_stats(struct iface *);
268 static void iface_refresh_stats(struct iface *);
269 static void iface_refresh_status(struct iface *);
270 static bool iface_is_synthetic(const struct iface *);
271 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
275 * This is deprecated. It is only for compatibility with broken device drivers
276 * in old versions of Linux that do not properly support VLANs when VLAN
277 * devices are not used. When broken device drivers are no longer in
278 * widespread use, we will delete these interfaces. */
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere;
283 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
284 static unsigned long int *collect_splinter_vlans(
285 const struct ovsrec_open_vswitch *);
286 static void configure_splinter_port(struct port *);
287 static void add_vlan_splinter_ports(struct bridge *,
288 const unsigned long int *splinter_vlans,
289 struct shash *ports);
292 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
294 struct shash iface_hints;
295 static bool initialized = false;
302 shash_init(&iface_hints);
305 for (i = 0; i < cfg->n_bridges; i++) {
306 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
309 for (j = 0; j < br_cfg->n_ports; j++) {
310 struct ovsrec_port *port_cfg = br_cfg->ports[j];
313 for (k = 0; k < port_cfg->n_interfaces; k++) {
314 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
315 struct iface_hint *iface_hint;
317 iface_hint = xmalloc(sizeof *iface_hint);
318 iface_hint->br_name = br_cfg->name;
319 iface_hint->br_type = br_cfg->datapath_type;
320 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
322 shash_add(&iface_hints, if_cfg->name, iface_hint);
328 ofproto_init(&iface_hints);
330 shash_destroy_free_data(&iface_hints);
334 /* Public functions. */
336 /* Initializes the bridge module, configuring it to obtain its configuration
337 * from an OVSDB server accessed over 'remote', which should be a string in a
338 * form acceptable to ovsdb_idl_create(). */
340 bridge_init(const char *remote)
342 /* Create connection to database. */
343 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
344 idl_seqno = ovsdb_idl_get_seqno(idl);
345 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
346 ovsdb_idl_verify_write_only(idl);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
349 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
354 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
357 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
358 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
361 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
363 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
383 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
384 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
387 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
388 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
389 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
393 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
396 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
398 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
401 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
406 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
407 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
409 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
411 /* Register unixctl commands. */
412 unixctl_command_register("qos/show", "interface", 1, 1,
413 qos_unixctl_show, NULL);
414 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
415 bridge_unixctl_dump_flows, NULL);
416 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
417 bridge_unixctl_reconnect, NULL);
427 struct bridge *br, *next_br;
429 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
432 ovsdb_idl_destroy(idl);
435 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
436 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
437 * responsible for freeing '*managersp' (with free()).
439 * You may be asking yourself "why does ovs-vswitchd care?", because
440 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
441 * should not be and in fact is not directly involved in that. But
442 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
443 * it has to tell in-band control where the managers are to enable that.
444 * (Thus, only managers connected in-band are collected.)
447 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
448 struct sockaddr_in **managersp, size_t *n_managersp)
450 struct sockaddr_in *managers = NULL;
451 size_t n_managers = 0;
455 /* Collect all of the potential targets from the "targets" columns of the
456 * rows pointed to by "manager_options", excluding any that are
459 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
460 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
462 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
463 sset_find_and_delete(&targets, m->target);
465 sset_add(&targets, m->target);
469 /* Now extract the targets' IP addresses. */
470 if (!sset_is_empty(&targets)) {
473 managers = xmalloc(sset_count(&targets) * sizeof *managers);
474 SSET_FOR_EACH (target, &targets) {
475 struct sockaddr_in *sin = &managers[n_managers];
477 if (stream_parse_target_with_default_port(target,
484 sset_destroy(&targets);
486 *managersp = managers;
487 *n_managersp = n_managers;
491 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
493 unsigned long int *splinter_vlans;
496 COVERAGE_INC(bridge_reconfigure);
498 ovs_assert(!reconfiguring);
499 reconfiguring = true;
501 ofproto_set_flow_eviction_threshold(
502 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
503 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
505 ofproto_set_n_handler_threads(
506 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0));
508 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
509 "force-miss-model"));
511 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
512 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
513 * configuration otherwise.
515 * This is mostly an update to bridge data structures. Nothing is pushed
516 * down to ofproto or lower layers. */
517 add_del_bridges(ovs_cfg);
518 splinter_vlans = collect_splinter_vlans(ovs_cfg);
519 HMAP_FOR_EACH (br, node, &all_bridges) {
520 bridge_add_del_ports(br, splinter_vlans);
522 free(splinter_vlans);
524 /* Delete datapaths that are no longer configured, and create ones which
525 * don't exist but should. */
526 bridge_update_ofprotos();
528 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
529 HMAP_FOR_EACH (br, node, &all_bridges) {
530 bridge_refresh_ofp_port(br);
533 /* Clear database records for "if_cfg"s which haven't been instantiated. */
534 HMAP_FOR_EACH (br, node, &all_bridges) {
535 struct if_cfg *if_cfg;
537 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
538 iface_clear_db_record(if_cfg->cfg);
542 reconfigure_system_stats(ovs_cfg);
546 bridge_reconfigure_ofp(void)
548 long long int deadline;
551 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
553 /* The kernel will reject any attempt to add a given port to a datapath if
554 * that port already belongs to a different datapath, so we must do all
555 * port deletions before any port additions. */
556 HMAP_FOR_EACH (br, node, &all_bridges) {
557 struct ofpp_garbage *garbage, *next;
559 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
560 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
561 * internal datastructures may not be consistent. Eventually, when
562 * port additions and deletions are cheaper, these calls should be
565 ofproto_port_del(br->ofproto, garbage->ofp_port);
566 list_remove(&garbage->list_node);
569 if (time_msec() >= deadline) {
576 HMAP_FOR_EACH (br, node, &all_bridges) {
577 struct if_cfg *if_cfg, *next;
579 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
580 iface_create(br, if_cfg, OFPP_NONE);
581 if (time_msec() >= deadline) {
591 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
593 struct sockaddr_in *managers;
594 int sflow_bridge_number;
599 ovs_assert(reconfiguring);
600 done = bridge_reconfigure_ofp();
602 /* Complete the configuration. */
603 sflow_bridge_number = 0;
604 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
605 HMAP_FOR_EACH (br, node, &all_bridges) {
608 /* We need the datapath ID early to allow LACP ports to use it as the
609 * default system ID. */
610 bridge_configure_datapath_id(br);
612 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
615 port_configure(port);
617 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
618 iface_configure_cfm(iface);
619 iface_configure_qos(iface, port->cfg->qos);
620 iface_set_mac(iface);
621 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
625 bridge_configure_mirrors(br);
626 bridge_configure_forward_bpdu(br);
627 bridge_configure_mac_table(br);
628 bridge_configure_remotes(br, managers, n_managers);
629 bridge_configure_netflow(br);
630 bridge_configure_sflow(br, &sflow_bridge_number);
631 bridge_configure_ipfix(br);
632 bridge_configure_stp(br);
633 bridge_configure_tables(br);
634 bridge_configure_dp_desc(br);
636 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
637 /* XXX: Remove this warning message eventually. */
638 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
639 " moved to the Open_vSwitch table. Ignoring its"
640 " setting in the bridge table.");
648 /* Delete ofprotos which aren't configured or have the wrong type. Create
649 * ofprotos which don't exist but need to. */
651 bridge_update_ofprotos(void)
653 struct bridge *br, *next;
658 /* Delete ofprotos with no bridge or with the wrong type. */
661 ofproto_enumerate_types(&types);
662 SSET_FOR_EACH (type, &types) {
665 ofproto_enumerate_names(type, &names);
666 SSET_FOR_EACH (name, &names) {
667 br = bridge_lookup(name);
668 if (!br || strcmp(type, br->type)) {
669 ofproto_delete(name, type);
673 sset_destroy(&names);
674 sset_destroy(&types);
676 /* Add ofprotos for bridges which don't have one yet. */
677 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
685 /* Remove ports from any datapath with the same name as 'br'. If we
686 * don't do this, creating 'br''s ofproto will fail because a port with
687 * the same name as its local port already exists. */
688 HMAP_FOR_EACH (br2, node, &all_bridges) {
689 struct ofproto_port ofproto_port;
695 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
697 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
699 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
700 ovs_strerror(error));
702 ofproto_port_destroy(&ofproto_port);
706 error = ofproto_create(br->name, br->type, &br->ofproto);
708 VLOG_ERR("failed to create bridge %s: %s", br->name,
709 ovs_strerror(error));
716 port_configure(struct port *port)
718 const struct ovsrec_port *cfg = port->cfg;
719 struct bond_settings bond_settings;
720 struct lacp_settings lacp_settings;
721 struct ofproto_bundle_settings s;
724 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
725 configure_splinter_port(port);
734 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
735 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
736 s.slaves[s.n_slaves++] = iface->ofp_port;
741 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
745 /* Get VLAN trunks. */
748 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
752 if (cfg->vlan_mode) {
753 if (!strcmp(cfg->vlan_mode, "access")) {
754 s.vlan_mode = PORT_VLAN_ACCESS;
755 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
756 s.vlan_mode = PORT_VLAN_TRUNK;
757 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
758 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
759 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
760 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
762 /* This "can't happen" because ovsdb-server should prevent it. */
763 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
764 s.vlan_mode = PORT_VLAN_TRUNK;
768 s.vlan_mode = PORT_VLAN_ACCESS;
770 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
774 s.vlan_mode = PORT_VLAN_TRUNK;
777 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
780 /* Get LACP settings. */
781 s.lacp = port_configure_lacp(port, &lacp_settings);
785 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
786 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
787 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
790 s.lacp_slaves = NULL;
793 /* Get bond settings. */
794 if (s.n_slaves > 1) {
795 s.bond = &bond_settings;
796 port_configure_bond(port, &bond_settings);
799 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
800 netdev_set_miimon_interval(iface->netdev, 0);
805 ofproto_bundle_register(port->bridge->ofproto, port, &s);
814 bridge_configure_flow_miss_model(const char *opt)
816 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
819 if (strcmp(opt, "with-facets")) {
820 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
821 } else if (strcmp(opt, "without-facets")) {
822 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
826 ofproto_set_flow_miss_model(model);
829 /* Pick local port hardware address and datapath ID for 'br'. */
831 bridge_configure_datapath_id(struct bridge *br)
833 uint8_t ea[ETH_ADDR_LEN];
835 struct iface *local_iface;
836 struct iface *hw_addr_iface;
839 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
840 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
842 int error = netdev_set_etheraddr(local_iface->netdev, ea);
844 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
845 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
846 "Ethernet address: %s",
847 br->name, ovs_strerror(error));
850 memcpy(br->ea, ea, ETH_ADDR_LEN);
852 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
853 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
854 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
855 ofproto_set_datapath_id(br->ofproto, dpid);
858 dpid_string = xasprintf("%016"PRIx64, dpid);
859 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
863 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
866 bridge_get_allowed_versions(struct bridge *br)
868 if (!br->cfg->n_protocols)
871 return ofputil_versions_from_strings(br->cfg->protocols,
872 br->cfg->n_protocols);
875 /* Set NetFlow configuration on 'br'. */
877 bridge_configure_netflow(struct bridge *br)
879 struct ovsrec_netflow *cfg = br->cfg->netflow;
880 struct netflow_options opts;
883 ofproto_set_netflow(br->ofproto, NULL);
887 memset(&opts, 0, sizeof opts);
889 /* Get default NetFlow configuration from datapath.
890 * Apply overrides from 'cfg'. */
891 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
892 if (cfg->engine_type) {
893 opts.engine_type = *cfg->engine_type;
895 if (cfg->engine_id) {
896 opts.engine_id = *cfg->engine_id;
899 /* Configure active timeout interval. */
900 opts.active_timeout = cfg->active_timeout;
901 if (!opts.active_timeout) {
902 opts.active_timeout = -1;
903 } else if (opts.active_timeout < 0) {
904 VLOG_WARN("bridge %s: active timeout interval set to negative "
905 "value, using default instead (%d seconds)", br->name,
906 NF_ACTIVE_TIMEOUT_DEFAULT);
907 opts.active_timeout = -1;
910 /* Add engine ID to interface number to disambiguate bridgs? */
911 opts.add_id_to_iface = cfg->add_id_to_interface;
912 if (opts.add_id_to_iface) {
913 if (opts.engine_id > 0x7f) {
914 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
915 "another vswitch, choose an engine id less than 128",
918 if (hmap_count(&br->ports) > 508) {
919 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
920 "another port when more than 508 ports are used",
926 sset_init(&opts.collectors);
927 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
930 if (ofproto_set_netflow(br->ofproto, &opts)) {
931 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
933 sset_destroy(&opts.collectors);
936 /* Set sFlow configuration on 'br'. */
938 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
940 const struct ovsrec_sflow *cfg = br->cfg->sflow;
941 struct ovsrec_controller **controllers;
942 struct ofproto_sflow_options oso;
943 size_t n_controllers;
947 ofproto_set_sflow(br->ofproto, NULL);
951 memset(&oso, 0, sizeof oso);
953 sset_init(&oso.targets);
954 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
956 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
958 oso.sampling_rate = *cfg->sampling;
961 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
963 oso.polling_interval = *cfg->polling;
966 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
968 oso.header_len = *cfg->header;
971 oso.sub_id = (*sflow_bridge_number)++;
972 oso.agent_device = cfg->agent;
974 oso.control_ip = NULL;
975 n_controllers = bridge_get_controllers(br, &controllers);
976 for (i = 0; i < n_controllers; i++) {
977 if (controllers[i]->local_ip) {
978 oso.control_ip = controllers[i]->local_ip;
982 ofproto_set_sflow(br->ofproto, &oso);
984 sset_destroy(&oso.targets);
987 /* Returns whether a IPFIX row is valid. */
989 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
991 return ipfix && ipfix->n_targets > 0;
994 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
996 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
997 const struct bridge *br)
999 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1002 /* Set IPFIX configuration on 'br'. */
1004 bridge_configure_ipfix(struct bridge *br)
1006 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1007 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1008 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1009 struct ofproto_ipfix_bridge_exporter_options be_opts;
1010 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1011 size_t n_fe_opts = 0;
1013 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1014 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1019 if (!valid_be_cfg && n_fe_opts == 0) {
1020 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1025 memset(&be_opts, 0, sizeof be_opts);
1027 sset_init(&be_opts.targets);
1028 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1030 if (be_cfg->sampling) {
1031 be_opts.sampling_rate = *be_cfg->sampling;
1033 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1035 if (be_cfg->obs_domain_id) {
1036 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1038 if (be_cfg->obs_point_id) {
1039 be_opts.obs_point_id = *be_cfg->obs_point_id;
1041 if (be_cfg->cache_active_timeout) {
1042 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1044 if (be_cfg->cache_max_flows) {
1045 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1049 if (n_fe_opts > 0) {
1050 struct ofproto_ipfix_flow_exporter_options *opts;
1051 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1053 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1054 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1055 opts->collector_set_id = fe_cfg->id;
1056 sset_init(&opts->targets);
1057 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1058 fe_cfg->ipfix->n_targets);
1059 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1060 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1061 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1062 ? *fe_cfg->ipfix->cache_max_flows : 0;
1068 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1072 sset_destroy(&be_opts.targets);
1075 if (n_fe_opts > 0) {
1076 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1078 for (i = 0; i < n_fe_opts; i++) {
1079 sset_destroy(&opts->targets);
1087 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1088 struct ofproto_port_stp_settings *port_s,
1089 int *port_num_counter, unsigned long *port_num_bitmap)
1091 const char *config_str;
1092 struct iface *iface;
1094 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1095 port_s->enable = false;
1098 port_s->enable = true;
1101 /* STP over bonds is not supported. */
1102 if (!list_is_singleton(&port->ifaces)) {
1103 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1105 port_s->enable = false;
1109 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1111 /* Internal ports shouldn't participate in spanning tree, so
1113 if (!strcmp(iface->type, "internal")) {
1114 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1115 port_s->enable = false;
1119 /* STP on mirror output ports is not supported. */
1120 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1121 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1122 port_s->enable = false;
1126 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1128 unsigned long int port_num = strtoul(config_str, NULL, 0);
1129 int port_idx = port_num - 1;
1131 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1132 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1133 port_s->enable = false;
1137 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1138 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1139 port->name, port_num);
1140 port_s->enable = false;
1143 bitmap_set1(port_num_bitmap, port_idx);
1144 port_s->port_num = port_idx;
1146 if (*port_num_counter >= STP_MAX_PORTS) {
1147 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1148 port_s->enable = false;
1152 port_s->port_num = (*port_num_counter)++;
1155 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1157 port_s->path_cost = strtoul(config_str, NULL, 10);
1159 enum netdev_features current;
1162 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1163 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1164 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1167 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1169 port_s->priority = strtoul(config_str, NULL, 0);
1171 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1175 /* Set spanning tree configuration on 'br'. */
1177 bridge_configure_stp(struct bridge *br)
1179 if (!br->cfg->stp_enable) {
1180 ofproto_set_stp(br->ofproto, NULL);
1182 struct ofproto_stp_settings br_s;
1183 const char *config_str;
1185 int port_num_counter;
1186 unsigned long *port_num_bitmap;
1188 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1190 uint8_t ea[ETH_ADDR_LEN];
1192 if (eth_addr_from_string(config_str, ea)) {
1193 br_s.system_id = eth_addr_to_uint64(ea);
1195 br_s.system_id = eth_addr_to_uint64(br->ea);
1196 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1197 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1200 br_s.system_id = eth_addr_to_uint64(br->ea);
1203 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1205 br_s.priority = strtoul(config_str, NULL, 0);
1207 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1210 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1212 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1214 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1217 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1219 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1221 br_s.max_age = STP_DEFAULT_MAX_AGE;
1224 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1226 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1228 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1231 /* Configure STP on the bridge. */
1232 if (ofproto_set_stp(br->ofproto, &br_s)) {
1233 VLOG_ERR("bridge %s: could not enable STP", br->name);
1237 /* Users must either set the port number with the "stp-port-num"
1238 * configuration on all ports or none. If manual configuration
1239 * is not done, then we allocate them sequentially. */
1240 port_num_counter = 0;
1241 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1242 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1243 struct ofproto_port_stp_settings port_s;
1244 struct iface *iface;
1246 port_configure_stp(br->ofproto, port, &port_s,
1247 &port_num_counter, port_num_bitmap);
1249 /* As bonds are not supported, just apply configuration to
1250 * all interfaces. */
1251 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1252 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1254 VLOG_ERR("port %s: could not enable STP", port->name);
1260 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1261 && port_num_counter) {
1262 VLOG_ERR("bridge %s: must manually configure all STP port "
1263 "IDs or none, disabling", br->name);
1264 ofproto_set_stp(br->ofproto, NULL);
1266 bitmap_free(port_num_bitmap);
1271 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1273 const struct port *port = port_lookup(br, name);
1274 return port && port_is_bond_fake_iface(port);
1278 port_is_bond_fake_iface(const struct port *port)
1280 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1284 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1286 struct bridge *br, *next;
1287 struct shash new_br;
1290 /* Collect new bridges' names and types. */
1291 shash_init(&new_br);
1292 for (i = 0; i < cfg->n_bridges; i++) {
1293 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1294 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1296 if (strchr(br_cfg->name, '/')) {
1297 /* Prevent remote ovsdb-server users from accessing arbitrary
1298 * directories, e.g. consider a bridge named "../../../etc/". */
1299 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1301 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1302 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1306 /* Get rid of deleted bridges or those whose types have changed.
1307 * Update 'cfg' of bridges that still exist. */
1308 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1309 br->cfg = shash_find_data(&new_br, br->name);
1310 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1311 br->cfg->datapath_type))) {
1316 /* Add new bridges. */
1317 for (i = 0; i < cfg->n_bridges; i++) {
1318 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1319 struct bridge *br = bridge_lookup(br_cfg->name);
1321 bridge_create(br_cfg);
1325 shash_destroy(&new_br);
1329 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1331 struct bridge *br = iface->port->bridge;
1333 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1334 iface->ofp_port = ofp_port;
1335 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1336 hash_ofp_port(ofp_port));
1337 iface_set_ofport(iface->cfg, ofp_port);
1340 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1341 * Returns 0 if successful, otherwise a positive errno value. */
1343 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1344 struct netdev *netdev)
1348 error = netdev_set_config(netdev, &iface_cfg->options);
1350 VLOG_WARN("could not configure network device %s (%s)",
1351 iface_cfg->name, ovs_strerror(error));
1356 /* This function determines whether 'ofproto_port', which is attached to
1357 * br->ofproto's datapath, is one that we want in 'br'.
1359 * If it is, it returns true, after creating an iface (if necessary),
1360 * configuring the iface's netdev according to the iface's options, and setting
1361 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1363 * If, on the other hand, 'port' should be removed, it returns false. The
1364 * caller should later detach the port from br->ofproto. */
1366 bridge_refresh_one_ofp_port(struct bridge *br,
1367 const struct ofproto_port *ofproto_port)
1369 const char *name = ofproto_port->name;
1370 const char *type = ofproto_port->type;
1371 ofp_port_t ofp_port = ofproto_port->ofp_port;
1373 struct iface *iface = iface_lookup(br, name);
1375 /* Check that the name-to-number mapping is one-to-one. */
1376 if (iface->ofp_port != OFPP_NONE) {
1377 VLOG_WARN("bridge %s: interface %s reported twice",
1380 } else if (iface_from_ofp_port(br, ofp_port)) {
1381 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1382 br->name, ofp_port);
1386 /* There's a configured interface named 'name'. */
1387 if (strcmp(type, iface->type)
1388 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1389 /* It's the wrong type, or it's the right type but can't be
1390 * configured as the user requested, so we must destroy it. */
1393 /* It's the right type and configured correctly. Keep it. */
1394 iface_set_ofp_port(iface, ofp_port);
1397 } else if (bridge_has_bond_fake_iface(br, name)
1398 && !strcmp(type, "internal")) {
1399 /* It's a bond fake iface. Keep it. */
1402 /* There's no configured interface named 'name', but there might be an
1403 * interface of that name queued to be created.
1405 * If there is, and it has the correct type, then try to configure it
1406 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1407 * delete it and later try to re-add it. */
1408 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1410 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1411 && iface_create(br, if_cfg, ofp_port));
1415 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1416 * consistent with the ofp_ports in "br->ofproto". */
1418 bridge_refresh_ofp_port(struct bridge *br)
1420 struct ofproto_port_dump dump;
1421 struct ofproto_port ofproto_port;
1422 struct port *port, *port_next;
1424 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1425 hmap_clear(&br->ifaces);
1426 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1427 struct iface *iface;
1429 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1430 iface->ofp_port = OFPP_NONE;
1434 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1435 * already exist in the datapath and promote them to full fledged "struct
1436 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1437 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1438 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1439 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1440 garbage->ofp_port = ofproto_port.ofp_port;
1441 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1445 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1446 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1447 * later they can be added to ofproto. */
1448 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1449 struct iface *iface, *iface_next;
1451 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1452 if (iface->ofp_port == OFPP_NONE) {
1453 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1454 iface_destroy(iface);
1458 if (list_is_empty(&port->ifaces)) {
1464 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1465 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1466 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1469 * If successful, returns 0 and stores the network device in '*netdevp'. On
1470 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1472 iface_do_create(const struct bridge *br,
1473 const struct if_cfg *if_cfg,
1474 ofp_port_t *ofp_portp, struct netdev **netdevp)
1476 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1477 const struct ovsrec_port *port_cfg = if_cfg->parent;
1478 struct netdev *netdev = NULL;
1481 if (netdev_is_reserved_name(iface_cfg->name)) {
1482 VLOG_WARN("could not create interface %s, name is reserved",
1488 error = netdev_open(iface_cfg->name,
1489 iface_get_type(iface_cfg, br->cfg), &netdev);
1491 VLOG_WARN("could not open network device %s (%s)",
1492 iface_cfg->name, ovs_strerror(error));
1496 error = iface_set_netdev_config(iface_cfg, netdev);
1501 if (*ofp_portp == OFPP_NONE) {
1502 ofp_port_t ofp_port = if_cfg->ofport;
1504 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1508 *ofp_portp = ofp_port;
1510 VLOG_INFO("bridge %s: added interface %s on port %d",
1511 br->name, iface_cfg->name, *ofp_portp);
1513 VLOG_DBG("bridge %s: interface %s is on port %d",
1514 br->name, iface_cfg->name, *ofp_portp);
1517 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1518 || iface_is_internal(iface_cfg, br->cfg)) {
1519 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1527 netdev_close(netdev);
1531 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1532 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1533 * automatically allocated for the iface. Takes ownership of and
1534 * deallocates 'if_cfg'.
1536 * Return true if an iface is successfully created, false otherwise. */
1538 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1540 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1541 const struct ovsrec_port *port_cfg = if_cfg->parent;
1543 struct netdev *netdev;
1544 struct iface *iface;
1549 /* Do the bits that can fail up front.
1551 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1552 * internal datastructures may not be consistent. Eventually, when port
1553 * additions and deletions are cheaper, these calls should be removed. */
1555 ovs_assert(!iface_lookup(br, iface_cfg->name));
1556 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1559 iface_set_ofport(iface_cfg, OFPP_NONE);
1560 iface_clear_db_record(iface_cfg);
1565 /* Get or create the port structure. */
1566 port = port_lookup(br, port_cfg->name);
1568 port = port_create(br, port_cfg);
1571 /* Create the iface structure. */
1572 iface = xzalloc(sizeof *iface);
1573 list_push_back(&port->ifaces, &iface->port_elem);
1574 hmap_insert(&br->iface_by_name, &iface->name_node,
1575 hash_string(iface_cfg->name, 0));
1577 iface->name = xstrdup(iface_cfg->name);
1578 iface->ofp_port = OFPP_NONE;
1579 iface->netdev = netdev;
1580 iface->type = iface_get_type(iface_cfg, br->cfg);
1581 iface->cfg = iface_cfg;
1583 iface_set_ofp_port(iface, ofp_port);
1585 /* Populate initial status in database. */
1586 iface_refresh_stats(iface);
1587 iface_refresh_status(iface);
1589 /* Add bond fake iface if necessary. */
1590 if (port_is_bond_fake_iface(port)) {
1591 struct ofproto_port ofproto_port;
1593 if (ofproto_port_query_by_name(br->ofproto, port->name,
1595 struct netdev *netdev;
1598 error = netdev_open(port->name, "internal", &netdev);
1600 ofp_port_t fake_ofp_port = if_cfg->ofport;
1602 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1603 netdev_close(netdev);
1605 VLOG_WARN("could not open network device %s (%s)",
1606 port->name, ovs_strerror(error));
1609 /* Already exists, nothing to do. */
1610 ofproto_port_destroy(&ofproto_port);
1615 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1621 /* Set forward BPDU option. */
1623 bridge_configure_forward_bpdu(struct bridge *br)
1625 ofproto_set_forward_bpdu(br->ofproto,
1626 smap_get_bool(&br->cfg->other_config,
1631 /* Set MAC learning table configuration for 'br'. */
1633 bridge_configure_mac_table(struct bridge *br)
1635 const char *idle_time_str;
1638 const char *mac_table_size_str;
1641 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1642 idle_time = (idle_time_str && atoi(idle_time_str)
1643 ? atoi(idle_time_str)
1644 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1646 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1647 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1648 ? atoi(mac_table_size_str)
1651 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1655 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1656 struct iface **hw_addr_iface)
1658 struct hmapx mirror_output_ports;
1661 bool found_addr = false;
1665 *hw_addr_iface = NULL;
1667 /* Did the user request a particular MAC? */
1668 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1669 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1670 if (eth_addr_is_multicast(ea)) {
1671 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1672 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1673 } else if (eth_addr_is_zero(ea)) {
1674 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1680 /* Mirror output ports don't participate in picking the local hardware
1681 * address. ofproto can't help us find out whether a given port is a
1682 * mirror output because we haven't configured mirrors yet, so we need to
1683 * accumulate them ourselves. */
1684 hmapx_init(&mirror_output_ports);
1685 for (i = 0; i < br->cfg->n_mirrors; i++) {
1686 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1687 if (m->output_port) {
1688 hmapx_add(&mirror_output_ports, m->output_port);
1692 /* Otherwise choose the minimum non-local MAC address among all of the
1694 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1695 uint8_t iface_ea[ETH_ADDR_LEN];
1696 struct iface *candidate;
1697 struct iface *iface;
1699 /* Mirror output ports don't participate. */
1700 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1704 /* Choose the MAC address to represent the port. */
1706 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1707 /* Find the interface with this Ethernet address (if any) so that
1708 * we can provide the correct devname to the caller. */
1709 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1710 uint8_t candidate_ea[ETH_ADDR_LEN];
1711 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1712 && eth_addr_equals(iface_ea, candidate_ea)) {
1717 /* Choose the interface whose MAC address will represent the port.
1718 * The Linux kernel bonding code always chooses the MAC address of
1719 * the first slave added to a bond, and the Fedora networking
1720 * scripts always add slaves to a bond in alphabetical order, so
1721 * for compatibility we choose the interface with the name that is
1722 * first in alphabetical order. */
1723 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1724 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1729 /* The local port doesn't count (since we're trying to choose its
1730 * MAC address anyway). */
1731 if (iface->ofp_port == OFPP_LOCAL) {
1736 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1742 /* Compare against our current choice. */
1743 if (!eth_addr_is_multicast(iface_ea) &&
1744 !eth_addr_is_local(iface_ea) &&
1745 !eth_addr_is_reserved(iface_ea) &&
1746 !eth_addr_is_zero(iface_ea) &&
1747 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1749 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1750 *hw_addr_iface = iface;
1756 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1757 *hw_addr_iface = NULL;
1760 hmapx_destroy(&mirror_output_ports);
1763 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1764 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1765 * an interface on 'br', then that interface must be passed in as
1766 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1767 * 'hw_addr_iface' must be passed in as a null pointer. */
1769 bridge_pick_datapath_id(struct bridge *br,
1770 const uint8_t bridge_ea[ETH_ADDR_LEN],
1771 struct iface *hw_addr_iface)
1774 * The procedure for choosing a bridge MAC address will, in the most
1775 * ordinary case, also choose a unique MAC that we can use as a datapath
1776 * ID. In some special cases, though, multiple bridges will end up with
1777 * the same MAC address. This is OK for the bridges, but it will confuse
1778 * the OpenFlow controller, because each datapath needs a unique datapath
1781 * Datapath IDs must be unique. It is also very desirable that they be
1782 * stable from one run to the next, so that policy set on a datapath
1785 const char *datapath_id;
1788 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1789 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1793 if (!hw_addr_iface) {
1795 * A purely internal bridge, that is, one that has no non-virtual
1796 * network devices on it at all, is difficult because it has no
1797 * natural unique identifier at all.
1799 * When the host is a XenServer, we handle this case by hashing the
1800 * host's UUID with the name of the bridge. Names of bridges are
1801 * persistent across XenServer reboots, although they can be reused if
1802 * an internal network is destroyed and then a new one is later
1803 * created, so this is fairly effective.
1805 * When the host is not a XenServer, we punt by using a random MAC
1806 * address on each run.
1808 const char *host_uuid = xenserver_get_host_uuid();
1810 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1811 dpid = dpid_from_hash(combined, strlen(combined));
1817 return eth_addr_to_uint64(bridge_ea);
1821 dpid_from_hash(const void *data, size_t n)
1823 uint8_t hash[SHA1_DIGEST_SIZE];
1825 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1826 sha1_bytes(data, n, hash);
1827 eth_addr_mark_random(hash);
1828 return eth_addr_to_uint64(hash);
1832 iface_refresh_status(struct iface *iface)
1836 enum netdev_features current;
1840 uint8_t mac[ETH_ADDR_LEN];
1844 if (iface_is_synthetic(iface)) {
1850 if (!netdev_get_status(iface->netdev, &smap)) {
1851 ovsrec_interface_set_status(iface->cfg, &smap);
1853 ovsrec_interface_set_status(iface->cfg, NULL);
1856 smap_destroy(&smap);
1858 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1859 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1861 ovsrec_interface_set_duplex(iface->cfg,
1862 netdev_features_is_full_duplex(current)
1864 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1866 ovsrec_interface_set_duplex(iface->cfg, NULL);
1867 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1870 error = netdev_get_mtu(iface->netdev, &mtu);
1873 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1875 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1878 error = netdev_get_etheraddr(iface->netdev, mac);
1880 char mac_string[32];
1882 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1883 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1885 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1888 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1889 * if there is no valid ifindex number. */
1890 ifindex64 = netdev_get_ifindex(iface->netdev);
1891 if (ifindex64 < 0) {
1894 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1897 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1900 iface_refresh_cfm_stats(struct iface *iface)
1902 const struct ovsrec_interface *cfg = iface->cfg;
1903 struct ofproto_cfm_status status;
1905 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1906 iface->ofp_port, &status)) {
1907 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1908 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1909 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1910 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1911 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1912 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1914 const char *reasons[CFM_FAULT_N_REASONS];
1915 int64_t cfm_health = status.health;
1916 int64_t cfm_flap_count = status.flap_count;
1917 bool faulted = status.faults != 0;
1920 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1923 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1924 int reason = 1 << i;
1925 if (status.faults & reason) {
1926 reasons[j++] = cfm_fault_reason_to_str(reason);
1929 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1931 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1933 if (status.remote_opstate >= 0) {
1934 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1935 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1937 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1940 ovsrec_interface_set_cfm_remote_mpids(cfg,
1941 (const int64_t *)status.rmps,
1943 if (cfm_health >= 0) {
1944 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1946 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1954 iface_refresh_stats(struct iface *iface)
1956 #define IFACE_STATS \
1957 IFACE_STAT(rx_packets, "rx_packets") \
1958 IFACE_STAT(tx_packets, "tx_packets") \
1959 IFACE_STAT(rx_bytes, "rx_bytes") \
1960 IFACE_STAT(tx_bytes, "tx_bytes") \
1961 IFACE_STAT(rx_dropped, "rx_dropped") \
1962 IFACE_STAT(tx_dropped, "tx_dropped") \
1963 IFACE_STAT(rx_errors, "rx_errors") \
1964 IFACE_STAT(tx_errors, "tx_errors") \
1965 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1966 IFACE_STAT(rx_over_errors, "rx_over_err") \
1967 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1968 IFACE_STAT(collisions, "collisions")
1970 #define IFACE_STAT(MEMBER, NAME) + 1
1971 enum { N_IFACE_STATS = IFACE_STATS };
1973 int64_t values[N_IFACE_STATS];
1974 char *keys[N_IFACE_STATS];
1977 struct netdev_stats stats;
1979 if (iface_is_synthetic(iface)) {
1983 /* Intentionally ignore return value, since errors will set 'stats' to
1984 * all-1s, and we will deal with that correctly below. */
1985 netdev_get_stats(iface->netdev, &stats);
1987 /* Copy statistics into keys[] and values[]. */
1989 #define IFACE_STAT(MEMBER, NAME) \
1990 if (stats.MEMBER != UINT64_MAX) { \
1992 values[n] = stats.MEMBER; \
1997 ovs_assert(n <= N_IFACE_STATS);
1999 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2004 br_refresh_stp_status(struct bridge *br)
2006 struct smap smap = SMAP_INITIALIZER(&smap);
2007 struct ofproto *ofproto = br->ofproto;
2008 struct ofproto_stp_status status;
2010 if (ofproto_get_stp_status(ofproto, &status)) {
2014 if (!status.enabled) {
2015 ovsrec_bridge_set_status(br->cfg, NULL);
2019 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2020 STP_ID_ARGS(status.bridge_id));
2021 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2022 STP_ID_ARGS(status.designated_root));
2023 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2025 ovsrec_bridge_set_status(br->cfg, &smap);
2026 smap_destroy(&smap);
2030 port_refresh_stp_status(struct port *port)
2032 struct ofproto *ofproto = port->bridge->ofproto;
2033 struct iface *iface;
2034 struct ofproto_port_stp_status status;
2036 int64_t int_values[3];
2039 if (port_is_synthetic(port)) {
2043 /* STP doesn't currently support bonds. */
2044 if (!list_is_singleton(&port->ifaces)) {
2045 ovsrec_port_set_status(port->cfg, NULL);
2049 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2051 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2055 if (!status.enabled) {
2056 ovsrec_port_set_status(port->cfg, NULL);
2057 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2061 /* Set Status column. */
2063 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2064 smap_add(&smap, "stp_state", stp_state_name(status.state));
2065 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2066 smap_add(&smap, "stp_role", stp_role_name(status.role));
2067 ovsrec_port_set_status(port->cfg, &smap);
2068 smap_destroy(&smap);
2070 /* Set Statistics column. */
2071 keys[0] = "stp_tx_count";
2072 int_values[0] = status.tx_count;
2073 keys[1] = "stp_rx_count";
2074 int_values[1] = status.rx_count;
2075 keys[2] = "stp_error_count";
2076 int_values[2] = status.error_count;
2078 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2079 ARRAY_SIZE(int_values));
2083 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2085 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2089 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2091 bool enable = enable_system_stats(cfg);
2093 system_stats_enable(enable);
2095 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2100 run_system_stats(void)
2102 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2105 stats = system_stats_run();
2107 struct ovsdb_idl_txn *txn;
2108 struct ovsdb_datum datum;
2110 txn = ovsdb_idl_txn_create(idl);
2111 ovsdb_datum_from_smap(&datum, stats);
2112 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2114 ovsdb_idl_txn_commit(txn);
2115 ovsdb_idl_txn_destroy(txn);
2121 static inline const char *
2122 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2125 case OFPCR12_ROLE_EQUAL:
2127 case OFPCR12_ROLE_MASTER:
2129 case OFPCR12_ROLE_SLAVE:
2131 case OFPCR12_ROLE_NOCHANGE:
2133 return "*** INVALID ROLE ***";
2138 refresh_controller_status(void)
2142 const struct ovsrec_controller *cfg;
2146 /* Accumulate status for controllers on all bridges. */
2147 HMAP_FOR_EACH (br, node, &all_bridges) {
2148 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2151 /* Update each controller in the database with current status. */
2152 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2153 struct ofproto_controller_info *cinfo =
2154 shash_find_data(&info, cfg->target);
2157 struct smap smap = SMAP_INITIALIZER(&smap);
2158 const char **values = cinfo->pairs.values;
2159 const char **keys = cinfo->pairs.keys;
2162 for (i = 0; i < cinfo->pairs.n; i++) {
2163 smap_add(&smap, keys[i], values[i]);
2166 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2167 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2169 ovsrec_controller_set_status(cfg, &smap);
2170 smap_destroy(&smap);
2172 ovsrec_controller_set_is_connected(cfg, false);
2173 ovsrec_controller_set_role(cfg, NULL);
2174 ovsrec_controller_set_status(cfg, NULL);
2178 ofproto_free_ofproto_controller_info(&info);
2183 * Some information in the database must be kept as up-to-date as possible to
2184 * allow controllers to respond rapidly to network outages. We call these
2185 * statistics "instant" stats.
2187 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2188 * assuming that they've changed. The only means we have to determine whether
2189 * they have changed are:
2191 * - Try to commit changes to the database. If nothing changed, then
2192 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2195 * - instant_stats_run() is called late in the run loop, after anything that
2196 * might change any of the instant stats.
2198 * We use these two facts together to avoid waking the process up every
2199 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2202 /* Minimum interval between writing updates to the instant stats to the
2204 #define INSTANT_INTERVAL_MSEC 100
2206 /* Current instant stats database transaction, NULL if there is no ongoing
2208 static struct ovsdb_idl_txn *instant_txn;
2210 /* Next time (in msec on monotonic clock) at which we will update the instant
2212 static long long int instant_next_txn = LLONG_MIN;
2214 /* True if the run loop has run since we last saw that the instant stats were
2215 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2216 * to refresh the instant stats. */
2217 static bool instant_stats_could_have_changed;
2220 instant_stats_run(void)
2222 enum ovsdb_idl_txn_status status;
2224 instant_stats_could_have_changed = true;
2229 if (time_msec() < instant_next_txn) {
2232 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2234 instant_txn = ovsdb_idl_txn_create(idl);
2235 HMAP_FOR_EACH (br, node, &all_bridges) {
2236 struct iface *iface;
2239 br_refresh_stp_status(br);
2241 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2242 port_refresh_stp_status(port);
2245 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2246 enum netdev_flags flags;
2248 const char *link_state;
2249 int64_t link_resets;
2252 if (iface_is_synthetic(iface)) {
2256 current = ofproto_port_is_lacp_current(br->ofproto,
2260 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2262 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2265 error = netdev_get_flags(iface->netdev, &flags);
2267 const char *state = flags & NETDEV_UP ? "up" : "down";
2268 ovsrec_interface_set_admin_state(iface->cfg, state);
2270 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2273 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2274 ovsrec_interface_set_link_state(iface->cfg, link_state);
2276 link_resets = netdev_get_carrier_resets(iface->netdev);
2277 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2279 iface_refresh_cfm_stats(iface);
2282 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2284 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2285 smap_destroy(&smap);
2290 status = ovsdb_idl_txn_commit(instant_txn);
2291 if (status != TXN_INCOMPLETE) {
2292 ovsdb_idl_txn_destroy(instant_txn);
2295 if (status == TXN_UNCHANGED) {
2296 instant_stats_could_have_changed = false;
2301 instant_stats_wait(void)
2304 ovsdb_idl_txn_wait(instant_txn);
2305 } else if (instant_stats_could_have_changed) {
2306 poll_timer_wait_until(instant_next_txn);
2310 /* Performs periodic activity required by bridges that needs to be done with
2311 * the least possible latency.
2313 * It makes sense to call this function a couple of times per poll loop, to
2314 * provide a significant performance boost on some benchmarks with ofprotos
2315 * that use the ofproto-dpif implementation. */
2317 bridge_run_fast(void)
2324 ofproto_enumerate_types(&types);
2325 SSET_FOR_EACH (type, &types) {
2326 ofproto_type_run_fast(type);
2328 sset_destroy(&types);
2330 HMAP_FOR_EACH (br, node, &all_bridges) {
2331 ofproto_run_fast(br->ofproto);
2338 static struct ovsrec_open_vswitch null_cfg;
2339 const struct ovsrec_open_vswitch *cfg;
2340 struct ovsdb_idl_txn *reconf_txn = NULL;
2344 bool vlan_splinters_changed;
2347 ovsrec_open_vswitch_init(&null_cfg);
2349 /* (Re)configure if necessary. */
2350 if (!reconfiguring) {
2353 if (ovsdb_idl_is_lock_contended(idl)) {
2354 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2355 struct bridge *br, *next_br;
2357 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2358 "disabling this process (pid %ld) until it goes away",
2359 (long int) getpid());
2361 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2364 /* Since we will not be running system_stats_run() in this process
2365 * with the current situation of multiple ovs-vswitchd daemons,
2366 * disable system stats collection. */
2367 system_stats_enable(false);
2369 } else if (!ovsdb_idl_has_lock(idl)) {
2373 cfg = ovsrec_open_vswitch_first(idl);
2375 /* Initialize the ofproto library. This only needs to run once, but
2376 * it must be done after the configuration is set. If the
2377 * initialization has already occurred, bridge_init_ofproto()
2378 * returns immediately. */
2379 bridge_init_ofproto(cfg);
2381 /* Once the value of flow-restore-wait is false, we no longer should
2382 * check its value from the database. */
2383 if (cfg && ofproto_get_flow_restore_wait()) {
2384 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2385 "flow-restore-wait", false));
2388 /* Let each datapath type do the work that it needs to do. */
2390 ofproto_enumerate_types(&types);
2391 SSET_FOR_EACH (type, &types) {
2392 ofproto_type_run(type);
2394 sset_destroy(&types);
2396 /* Let each bridge do the work that it needs to do. */
2397 HMAP_FOR_EACH (br, node, &all_bridges) {
2398 ofproto_run(br->ofproto);
2401 /* Re-configure SSL. We do this on every trip through the main loop,
2402 * instead of just when the database changes, because the contents of the
2403 * key and certificate files can change without the database changing.
2405 * We do this before bridge_reconfigure() because that function might
2406 * initiate SSL connections and thus requires SSL to be configured. */
2407 if (cfg && cfg->ssl) {
2408 const struct ovsrec_ssl *ssl = cfg->ssl;
2410 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2411 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2414 if (!reconfiguring) {
2415 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2416 * usage has changed. */
2417 vlan_splinters_changed = false;
2418 if (vlan_splinters_enabled_anywhere) {
2419 HMAP_FOR_EACH (br, node, &all_bridges) {
2420 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2421 vlan_splinters_changed = true;
2427 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2428 idl_seqno = ovsdb_idl_get_seqno(idl);
2430 reconf_txn = ovsdb_idl_txn_create(idl);
2431 bridge_reconfigure(cfg);
2433 /* We still need to reconfigure to avoid dangling pointers to
2434 * now-destroyed ovsrec structures inside bridge data. */
2435 bridge_reconfigure(&null_cfg);
2440 if (reconfiguring) {
2442 reconf_txn = ovsdb_idl_txn_create(idl);
2445 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2446 reconfiguring = false;
2449 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2452 /* If we are completing our initial configuration for this run
2453 * of ovs-vswitchd, then keep the transaction around to monitor
2454 * it for completion. */
2455 if (!initial_config_done) {
2456 initial_config_done = true;
2457 daemonize_txn = reconf_txn;
2464 ovsdb_idl_txn_commit(reconf_txn);
2465 ovsdb_idl_txn_destroy(reconf_txn);
2469 if (daemonize_txn) {
2470 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2471 if (status != TXN_INCOMPLETE) {
2472 ovsdb_idl_txn_destroy(daemonize_txn);
2473 daemonize_txn = NULL;
2475 /* ovs-vswitchd has completed initialization, so allow the
2476 * process that forked us to exit successfully. */
2477 daemonize_complete();
2479 vlog_enable_async();
2481 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2485 /* Refresh interface and mirror stats if necessary. */
2486 if (time_msec() >= iface_stats_timer) {
2488 struct ovsdb_idl_txn *txn;
2490 txn = ovsdb_idl_txn_create(idl);
2491 HMAP_FOR_EACH (br, node, &all_bridges) {
2495 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2496 struct iface *iface;
2498 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2499 iface_refresh_stats(iface);
2500 iface_refresh_status(iface);
2504 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2505 mirror_refresh_stats(m);
2509 refresh_controller_status();
2510 ovsdb_idl_txn_commit(txn);
2511 ovsdb_idl_txn_destroy(txn); /* XXX */
2514 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2518 instant_stats_run();
2527 ovsdb_idl_wait(idl);
2528 if (daemonize_txn) {
2529 ovsdb_idl_txn_wait(daemonize_txn);
2532 if (reconfiguring) {
2533 poll_immediate_wake();
2537 ofproto_enumerate_types(&types);
2538 SSET_FOR_EACH (type, &types) {
2539 ofproto_type_wait(type);
2541 sset_destroy(&types);
2543 if (!hmap_is_empty(&all_bridges)) {
2546 HMAP_FOR_EACH (br, node, &all_bridges) {
2547 ofproto_wait(br->ofproto);
2549 poll_timer_wait_until(iface_stats_timer);
2552 system_stats_wait();
2553 instant_stats_wait();
2556 /* Adds some memory usage statistics for bridges into 'usage', for use with
2557 * memory_report(). */
2559 bridge_get_memory_usage(struct simap *usage)
2563 HMAP_FOR_EACH (br, node, &all_bridges) {
2564 ofproto_get_memory_usage(br->ofproto, usage);
2568 /* QoS unixctl user interface functions. */
2570 struct qos_unixctl_show_cbdata {
2572 struct iface *iface;
2576 qos_unixctl_show_queue(unsigned int queue_id,
2577 const struct smap *details,
2578 struct iface *iface,
2581 struct netdev_queue_stats stats;
2582 struct smap_node *node;
2585 ds_put_cstr(ds, "\n");
2587 ds_put_format(ds, "Queue %u:\n", queue_id);
2589 ds_put_cstr(ds, "Default:\n");
2592 SMAP_FOR_EACH (node, details) {
2593 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2596 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2598 if (stats.tx_packets != UINT64_MAX) {
2599 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2602 if (stats.tx_bytes != UINT64_MAX) {
2603 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2606 if (stats.tx_errors != UINT64_MAX) {
2607 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2610 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2611 queue_id, ovs_strerror(error));
2616 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2617 const char *argv[], void *aux OVS_UNUSED)
2619 struct ds ds = DS_EMPTY_INITIALIZER;
2620 struct smap smap = SMAP_INITIALIZER(&smap);
2621 struct iface *iface;
2623 struct smap_node *node;
2625 iface = iface_find(argv[1]);
2627 unixctl_command_reply_error(conn, "no such interface");
2631 netdev_get_qos(iface->netdev, &type, &smap);
2633 if (*type != '\0') {
2634 struct netdev_queue_dump dump;
2635 struct smap details;
2636 unsigned int queue_id;
2638 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2640 SMAP_FOR_EACH (node, &smap) {
2641 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2644 smap_init(&details);
2645 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2646 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2648 smap_destroy(&details);
2650 unixctl_command_reply(conn, ds_cstr(&ds));
2652 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2653 unixctl_command_reply_error(conn, ds_cstr(&ds));
2656 smap_destroy(&smap);
2660 /* Bridge reconfiguration functions. */
2662 bridge_create(const struct ovsrec_bridge *br_cfg)
2666 ovs_assert(!bridge_lookup(br_cfg->name));
2667 br = xzalloc(sizeof *br);
2669 br->name = xstrdup(br_cfg->name);
2670 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2673 /* Derive the default Ethernet address from the bridge's UUID. This should
2674 * be unique and it will be stable between ovs-vswitchd runs. */
2675 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2676 eth_addr_mark_random(br->default_ea);
2678 hmap_init(&br->ports);
2679 hmap_init(&br->ifaces);
2680 hmap_init(&br->iface_by_name);
2681 hmap_init(&br->mirrors);
2683 hmap_init(&br->if_cfg_todo);
2684 list_init(&br->ofpp_garbage);
2686 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2690 bridge_destroy(struct bridge *br)
2693 struct mirror *mirror, *next_mirror;
2694 struct port *port, *next_port;
2695 struct if_cfg *if_cfg, *next_if_cfg;
2696 struct ofpp_garbage *garbage, *next_garbage;
2698 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2701 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2702 mirror_destroy(mirror);
2704 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2705 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2708 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2709 &br->ofpp_garbage) {
2710 list_remove(&garbage->list_node);
2714 hmap_remove(&all_bridges, &br->node);
2715 ofproto_destroy(br->ofproto);
2716 hmap_destroy(&br->ifaces);
2717 hmap_destroy(&br->ports);
2718 hmap_destroy(&br->iface_by_name);
2719 hmap_destroy(&br->mirrors);
2720 hmap_destroy(&br->if_cfg_todo);
2727 static struct bridge *
2728 bridge_lookup(const char *name)
2732 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2733 if (!strcmp(br->name, name)) {
2740 /* Handle requests for a listing of all flows known by the OpenFlow
2741 * stack, including those normally hidden. */
2743 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2744 const char *argv[], void *aux OVS_UNUSED)
2749 br = bridge_lookup(argv[1]);
2751 unixctl_command_reply_error(conn, "Unknown bridge");
2756 ofproto_get_all_flows(br->ofproto, &results);
2758 unixctl_command_reply(conn, ds_cstr(&results));
2759 ds_destroy(&results);
2762 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2763 * connections and reconnect. If BRIDGE is not specified, then all bridges
2764 * drop their controller connections and reconnect. */
2766 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2767 const char *argv[], void *aux OVS_UNUSED)
2771 br = bridge_lookup(argv[1]);
2773 unixctl_command_reply_error(conn, "Unknown bridge");
2776 ofproto_reconnect_controllers(br->ofproto);
2778 HMAP_FOR_EACH (br, node, &all_bridges) {
2779 ofproto_reconnect_controllers(br->ofproto);
2782 unixctl_command_reply(conn, NULL);
2786 bridge_get_controllers(const struct bridge *br,
2787 struct ovsrec_controller ***controllersp)
2789 struct ovsrec_controller **controllers;
2790 size_t n_controllers;
2792 controllers = br->cfg->controller;
2793 n_controllers = br->cfg->n_controller;
2795 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2801 *controllersp = controllers;
2803 return n_controllers;
2807 bridge_queue_if_cfg(struct bridge *br,
2808 const struct ovsrec_interface *cfg,
2809 const struct ovsrec_port *parent)
2811 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2814 if_cfg->parent = parent;
2815 if_cfg->ofport = iface_pick_ofport(cfg);
2816 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2817 hash_string(if_cfg->cfg->name, 0));
2820 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2821 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2822 * 'br' needs to complete its configuration. */
2824 bridge_add_del_ports(struct bridge *br,
2825 const unsigned long int *splinter_vlans)
2827 struct shash_node *port_node;
2828 struct port *port, *next;
2829 struct shash new_ports;
2832 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2834 /* Collect new ports. */
2835 shash_init(&new_ports);
2836 for (i = 0; i < br->cfg->n_ports; i++) {
2837 const char *name = br->cfg->ports[i]->name;
2838 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2839 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2843 if (bridge_get_controllers(br, NULL)
2844 && !shash_find(&new_ports, br->name)) {
2845 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2846 br->name, br->name);
2848 ovsrec_interface_init(&br->synth_local_iface);
2849 ovsrec_port_init(&br->synth_local_port);
2851 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2852 br->synth_local_port.n_interfaces = 1;
2853 br->synth_local_port.name = br->name;
2855 br->synth_local_iface.name = br->name;
2856 br->synth_local_iface.type = "internal";
2858 br->synth_local_ifacep = &br->synth_local_iface;
2860 shash_add(&new_ports, br->name, &br->synth_local_port);
2863 if (splinter_vlans) {
2864 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2867 /* Get rid of deleted ports.
2868 * Get rid of deleted interfaces on ports that still exist. */
2869 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2870 port->cfg = shash_find_data(&new_ports, port->name);
2874 port_del_ifaces(port);
2878 /* Update iface->cfg and iface->type in interfaces that still exist.
2879 * Add new interfaces to creation queue. */
2880 SHASH_FOR_EACH (port_node, &new_ports) {
2881 const struct ovsrec_port *port = port_node->data;
2884 for (i = 0; i < port->n_interfaces; i++) {
2885 const struct ovsrec_interface *cfg = port->interfaces[i];
2886 struct iface *iface = iface_lookup(br, cfg->name);
2887 const char *type = iface_get_type(cfg, br->cfg);
2892 } else if (!strcmp(type, "null")) {
2893 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2894 " may be removed in February 2013. Please email"
2895 " dev@openvswitch.org with concerns.",
2898 bridge_queue_if_cfg(br, cfg, port);
2903 shash_destroy(&new_ports);
2906 /* Initializes 'oc' appropriately as a management service controller for
2909 * The caller must free oc->target when it is no longer needed. */
2911 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2912 struct ofproto_controller *oc)
2914 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2915 oc->max_backoff = 0;
2916 oc->probe_interval = 60;
2917 oc->band = OFPROTO_OUT_OF_BAND;
2919 oc->burst_limit = 0;
2920 oc->enable_async_msgs = true;
2923 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2925 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2926 struct ofproto_controller *oc)
2930 oc->target = c->target;
2931 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2932 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2933 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2934 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2935 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2936 oc->burst_limit = (c->controller_burst_limit
2937 ? *c->controller_burst_limit : 0);
2938 oc->enable_async_msgs = (!c->enable_async_messages
2939 || *c->enable_async_messages);
2940 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2941 if (dscp < 0 || dscp > 63) {
2942 dscp = DSCP_DEFAULT;
2947 /* Configures the IP stack for 'br''s local interface properly according to the
2948 * configuration in 'c'. */
2950 bridge_configure_local_iface_netdev(struct bridge *br,
2951 struct ovsrec_controller *c)
2953 struct netdev *netdev;
2954 struct in_addr mask, gateway;
2956 struct iface *local_iface;
2959 /* If there's no local interface or no IP address, give up. */
2960 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2961 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2965 /* Bring up the local interface. */
2966 netdev = local_iface->netdev;
2967 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2969 /* Configure the IP address and netmask. */
2970 if (!c->local_netmask
2971 || !inet_aton(c->local_netmask, &mask)
2973 mask.s_addr = guess_netmask(ip.s_addr);
2975 if (!netdev_set_in4(netdev, ip, mask)) {
2976 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2977 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2980 /* Configure the default gateway. */
2981 if (c->local_gateway
2982 && inet_aton(c->local_gateway, &gateway)
2983 && gateway.s_addr) {
2984 if (!netdev_add_router(netdev, gateway)) {
2985 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2986 br->name, IP_ARGS(gateway.s_addr));
2991 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2992 * in either string are treated as equal to any number of slashes in the other,
2993 * e.g. "x///y" is equal to "x/y".
2995 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2996 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2997 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2998 * 'b' against a prefix of 'a'.
3001 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3003 const char *b_start = b;
3005 if (b - b_start >= b_stoplen) {
3007 } else if (*a != *b) {
3009 } else if (*a == '/') {
3010 a += strspn(a, "/");
3011 b += strspn(b, "/");
3012 } else if (*a == '\0') {
3022 bridge_configure_remotes(struct bridge *br,
3023 const struct sockaddr_in *managers, size_t n_managers)
3025 bool disable_in_band;
3027 struct ovsrec_controller **controllers;
3028 size_t n_controllers;
3030 enum ofproto_fail_mode fail_mode;
3032 struct ofproto_controller *ocs;
3036 /* Check if we should disable in-band control on this bridge. */
3037 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3040 /* Set OpenFlow queue ID for in-band control. */
3041 ofproto_set_in_band_queue(br->ofproto,
3042 smap_get_int(&br->cfg->other_config,
3043 "in-band-queue", -1));
3045 if (disable_in_band) {
3046 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3048 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3051 n_controllers = bridge_get_controllers(br, &controllers);
3053 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3056 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3057 for (i = 0; i < n_controllers; i++) {
3058 struct ovsrec_controller *c = controllers[i];
3060 if (!strncmp(c->target, "punix:", 6)
3061 || !strncmp(c->target, "unix:", 5)) {
3062 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3065 if (!strncmp(c->target, "unix:", 5)) {
3066 /* Connect to a listening socket */
3067 whitelist = xasprintf("unix:%s/", ovs_rundir());
3068 if (strchr(c->target, '/') &&
3069 !equal_pathnames(c->target, whitelist,
3070 strlen(whitelist))) {
3071 /* Absolute path specified, but not in ovs_rundir */
3072 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3073 "controller \"%s\" due to possibility for "
3074 "remote exploit. Instead, specify socket "
3075 "in whitelisted \"%s\" or connect to "
3076 "\"unix:%s/%s.mgmt\" (which is always "
3077 "available without special configuration).",
3078 br->name, c->target, whitelist,
3079 ovs_rundir(), br->name);
3084 whitelist = xasprintf("punix:%s/%s.controller",
3085 ovs_rundir(), br->name);
3086 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3087 /* Prevent remote ovsdb-server users from accessing
3088 * arbitrary Unix domain sockets and overwriting arbitrary
3090 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3091 "controller \"%s\" due to possibility of "
3092 "overwriting local files. Instead, specify "
3093 "whitelisted \"%s\" or connect to "
3094 "\"unix:%s/%s.mgmt\" (which is always "
3095 "available without special configuration).",
3096 br->name, c->target, whitelist,
3097 ovs_rundir(), br->name);
3106 bridge_configure_local_iface_netdev(br, c);
3107 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3108 if (disable_in_band) {
3109 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3114 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3115 bridge_get_allowed_versions(br));
3116 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3119 /* Set the fail-mode. */
3120 fail_mode = !br->cfg->fail_mode
3121 || !strcmp(br->cfg->fail_mode, "standalone")
3122 ? OFPROTO_FAIL_STANDALONE
3123 : OFPROTO_FAIL_SECURE;
3124 ofproto_set_fail_mode(br->ofproto, fail_mode);
3126 /* Configure OpenFlow controller connection snooping. */
3127 if (!ofproto_has_snoops(br->ofproto)) {
3131 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3132 ovs_rundir(), br->name));
3133 ofproto_set_snoops(br->ofproto, &snoops);
3134 sset_destroy(&snoops);
3139 bridge_configure_tables(struct bridge *br)
3141 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3145 n_tables = ofproto_get_n_tables(br->ofproto);
3147 for (i = 0; i < n_tables; i++) {
3148 struct ofproto_table_settings s;
3151 s.max_flows = UINT_MAX;
3155 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3156 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3159 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3160 s.max_flows = *cfg->flow_limit;
3162 if (cfg->overflow_policy
3163 && !strcmp(cfg->overflow_policy, "evict")) {
3166 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3167 for (k = 0; k < cfg->n_groups; k++) {
3168 const char *string = cfg->groups[k];
3171 msg = mf_parse_subfield__(&s.groups[k], &string);
3173 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3174 "'groups' (%s)", br->name, i, msg);
3176 } else if (*string) {
3177 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3178 "element '%s' contains trailing garbage",
3179 br->name, i, cfg->groups[k]);
3187 ofproto_configure_table(br->ofproto, i, &s);
3191 for (; j < br->cfg->n_flow_tables; j++) {
3192 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3193 "%"PRId64" not supported by this datapath", br->name,
3194 br->cfg->key_flow_tables[j]);
3199 bridge_configure_dp_desc(struct bridge *br)
3201 ofproto_set_dp_desc(br->ofproto,
3202 smap_get(&br->cfg->other_config, "dp-desc"));
3205 /* Port functions. */
3207 static struct port *
3208 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3212 port = xzalloc(sizeof *port);
3214 port->name = xstrdup(cfg->name);
3216 list_init(&port->ifaces);
3218 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3222 /* Deletes interfaces from 'port' that are no longer configured for it. */
3224 port_del_ifaces(struct port *port)
3226 struct iface *iface, *next;
3227 struct sset new_ifaces;
3230 /* Collect list of new interfaces. */
3231 sset_init(&new_ifaces);
3232 for (i = 0; i < port->cfg->n_interfaces; i++) {
3233 const char *name = port->cfg->interfaces[i]->name;
3234 const char *type = port->cfg->interfaces[i]->type;
3235 if (strcmp(type, "null")) {
3236 sset_add(&new_ifaces, name);
3240 /* Get rid of deleted interfaces. */
3241 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3242 if (!sset_contains(&new_ifaces, iface->name)) {
3243 iface_destroy(iface);
3247 sset_destroy(&new_ifaces);
3251 port_destroy(struct port *port)
3254 struct bridge *br = port->bridge;
3255 struct iface *iface, *next;
3258 ofproto_bundle_unregister(br->ofproto, port);
3261 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3262 iface_destroy(iface);
3265 hmap_remove(&br->ports, &port->hmap_node);
3271 static struct port *
3272 port_lookup(const struct bridge *br, const char *name)
3276 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3278 if (!strcmp(port->name, name)) {
3286 enable_lacp(struct port *port, bool *activep)
3288 if (!port->cfg->lacp) {
3289 /* XXX when LACP implementation has been sufficiently tested, enable by
3290 * default and make active on bonded ports. */
3292 } else if (!strcmp(port->cfg->lacp, "off")) {
3294 } else if (!strcmp(port->cfg->lacp, "active")) {
3297 } else if (!strcmp(port->cfg->lacp, "passive")) {
3301 VLOG_WARN("port %s: unknown LACP mode %s",
3302 port->name, port->cfg->lacp);
3307 static struct lacp_settings *
3308 port_configure_lacp(struct port *port, struct lacp_settings *s)
3310 const char *lacp_time, *system_id;
3313 if (!enable_lacp(port, &s->active)) {
3317 s->name = port->name;
3319 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3321 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3322 ETH_ADDR_SCAN_ARGS(s->id))) {
3323 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3324 " address.", port->name, system_id);
3328 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3331 if (eth_addr_is_zero(s->id)) {
3332 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3336 /* Prefer bondable links if unspecified. */
3337 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3339 s->priority = (priority > 0 && priority <= UINT16_MAX
3341 : UINT16_MAX - !list_is_short(&port->ifaces));
3343 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3344 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3346 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3347 "lacp-fallback-ab", false);
3353 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3355 int priority, portid, key;
3357 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3358 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3360 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3362 if (portid <= 0 || portid > UINT16_MAX) {
3363 portid = ofp_to_u16(iface->ofp_port);
3366 if (priority <= 0 || priority > UINT16_MAX) {
3367 priority = UINT16_MAX;
3370 if (key < 0 || key > UINT16_MAX) {
3374 s->name = iface->name;
3376 s->priority = priority;
3381 port_configure_bond(struct port *port, struct bond_settings *s)
3383 const char *detect_s;
3384 struct iface *iface;
3385 int miimon_interval;
3387 s->name = port->name;
3389 if (port->cfg->bond_mode) {
3390 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3391 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3392 port->name, port->cfg->bond_mode,
3393 bond_mode_to_string(s->balance));
3396 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3398 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3399 * active-backup. At some point we should remove this warning. */
3400 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3401 " in previous versions, the default bond_mode was"
3402 " balance-slb", port->name,
3403 bond_mode_to_string(s->balance));
3405 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3406 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3407 "please use another bond type or disable flood_vlans",
3411 miimon_interval = smap_get_int(&port->cfg->other_config,
3412 "bond-miimon-interval", 0);
3413 if (miimon_interval <= 0) {
3414 miimon_interval = 200;
3417 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3418 if (!detect_s || !strcmp(detect_s, "carrier")) {
3419 miimon_interval = 0;
3420 } else if (strcmp(detect_s, "miimon")) {
3421 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3422 "defaulting to carrier", port->name, detect_s);
3423 miimon_interval = 0;
3426 s->up_delay = MAX(0, port->cfg->bond_updelay);
3427 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3428 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3429 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3430 "bond-rebalance-interval", 10000);
3431 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3432 s->rebalance_interval = 1000;
3435 s->fake_iface = port->cfg->bond_fake_iface;
3437 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3438 "lacp-fallback-ab", false);
3440 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3441 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3445 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3446 * instead of obtaining it from the database. */
3448 port_is_synthetic(const struct port *port)
3450 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3453 /* Interface functions. */
3456 iface_is_internal(const struct ovsrec_interface *iface,
3457 const struct ovsrec_bridge *br)
3459 /* The local port and "internal" ports are always "internal". */
3460 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3463 /* Returns the correct network device type for interface 'iface' in bridge
3466 iface_get_type(const struct ovsrec_interface *iface,
3467 const struct ovsrec_bridge *br)
3471 /* The local port always has type "internal". Other ports take
3472 * their type from the database and default to "system" if none is
3474 if (iface_is_internal(iface, br)) {
3477 type = iface->type[0] ? iface->type : "system";
3480 return ofproto_port_open_type(br->datapath_type, type);
3484 iface_destroy(struct iface *iface)
3487 struct port *port = iface->port;
3488 struct bridge *br = port->bridge;
3490 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3491 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3494 if (iface->ofp_port != OFPP_NONE) {
3495 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3498 list_remove(&iface->port_elem);
3499 hmap_remove(&br->iface_by_name, &iface->name_node);
3501 netdev_close(iface->netdev);
3508 static struct iface *
3509 iface_lookup(const struct bridge *br, const char *name)
3511 struct iface *iface;
3513 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3514 &br->iface_by_name) {
3515 if (!strcmp(iface->name, name)) {
3523 static struct iface *
3524 iface_find(const char *name)
3526 const struct bridge *br;
3528 HMAP_FOR_EACH (br, node, &all_bridges) {
3529 struct iface *iface = iface_lookup(br, name);
3538 static struct if_cfg *
3539 if_cfg_lookup(const struct bridge *br, const char *name)
3541 struct if_cfg *if_cfg;
3543 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3545 if (!strcmp(if_cfg->cfg->name, name)) {
3553 static struct iface *
3554 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3556 struct iface *iface;
3558 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3560 if (iface->ofp_port == ofp_port) {
3567 /* Set Ethernet address of 'iface', if one is specified in the configuration
3570 iface_set_mac(struct iface *iface)
3572 uint8_t ea[ETH_ADDR_LEN];
3574 if (!strcmp(iface->type, "internal")
3575 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3576 if (iface->ofp_port == OFPP_LOCAL) {
3577 VLOG_ERR("interface %s: ignoring mac in Interface record "
3578 "(use Bridge record to set local port's mac)",
3580 } else if (eth_addr_is_multicast(ea)) {
3581 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3584 int error = netdev_set_etheraddr(iface->netdev, ea);
3586 VLOG_ERR("interface %s: setting MAC failed (%s)",
3587 iface->name, ovs_strerror(error));
3593 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3595 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3598 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3599 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3600 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3604 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3605 * sets the "ofport" field to -1.
3607 * This is appropriate when 'if_cfg''s interface cannot be created or is
3608 * otherwise invalid. */
3610 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3612 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3613 ovsrec_interface_set_status(if_cfg, NULL);
3614 ovsrec_interface_set_admin_state(if_cfg, NULL);
3615 ovsrec_interface_set_duplex(if_cfg, NULL);
3616 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3617 ovsrec_interface_set_link_state(if_cfg, NULL);
3618 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3619 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3620 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3621 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3622 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3623 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3624 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3625 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3630 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3632 union ovsdb_atom atom;
3634 atom.integer = target;
3635 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3639 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3641 struct ofpbuf queues_buf;
3643 ofpbuf_init(&queues_buf, 0);
3645 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3646 netdev_set_qos(iface->netdev, NULL, NULL);
3648 const struct ovsdb_datum *queues;
3649 struct netdev_queue_dump dump;
3650 unsigned int queue_id;
3651 struct smap details;
3655 /* Configure top-level Qos for 'iface'. */
3656 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3658 /* Deconfigure queues that were deleted. */
3659 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3661 smap_init(&details);
3662 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3663 if (!queue_ids_include(queues, queue_id)) {
3664 netdev_delete_queue(iface->netdev, queue_id);
3667 smap_destroy(&details);
3669 /* Configure queues for 'iface'. */
3671 for (i = 0; i < qos->n_queues; i++) {
3672 const struct ovsrec_queue *queue = qos->value_queues[i];
3673 unsigned int queue_id = qos->key_queues[i];
3675 if (queue_id == 0) {
3679 if (queue->n_dscp == 1) {
3680 struct ofproto_port_queue *port_queue;
3682 port_queue = ofpbuf_put_uninit(&queues_buf,
3683 sizeof *port_queue);
3684 port_queue->queue = queue_id;
3685 port_queue->dscp = queue->dscp[0];
3688 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3691 struct smap details;
3693 smap_init(&details);
3694 netdev_set_queue(iface->netdev, 0, &details);
3695 smap_destroy(&details);
3699 if (iface->ofp_port != OFPP_NONE) {
3700 const struct ofproto_port_queue *port_queues = queues_buf.data;
3701 size_t n_queues = queues_buf.size / sizeof *port_queues;
3703 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3704 port_queues, n_queues);
3707 netdev_set_policing(iface->netdev,
3708 iface->cfg->ingress_policing_rate,
3709 iface->cfg->ingress_policing_burst);
3711 ofpbuf_uninit(&queues_buf);
3715 iface_configure_cfm(struct iface *iface)
3717 const struct ovsrec_interface *cfg = iface->cfg;
3718 const char *opstate_str;
3719 const char *cfm_ccm_vlan;
3720 struct cfm_settings s;
3721 struct smap netdev_args;
3723 if (!cfg->n_cfm_mpid) {
3724 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3728 s.check_tnl_key = false;
3729 smap_init(&netdev_args);
3730 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3731 const char *key = smap_get(&netdev_args, "key");
3732 const char *in_key = smap_get(&netdev_args, "in_key");
3734 s.check_tnl_key = (key && !strcmp(key, "flow"))
3735 || (in_key && !strcmp(in_key, "flow"));
3737 smap_destroy(&netdev_args);
3739 s.mpid = *cfg->cfm_mpid;
3740 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3741 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3742 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3744 if (s.interval <= 0) {
3748 if (!cfm_ccm_vlan) {
3750 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3751 s.ccm_vlan = CFM_RANDOM_VLAN;
3753 s.ccm_vlan = atoi(cfm_ccm_vlan);
3754 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3759 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3761 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3763 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3764 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3766 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3769 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3770 * instead of obtaining it from the database. */
3772 iface_is_synthetic(const struct iface *iface)
3774 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3778 iface_pick_ofport(const struct ovsrec_interface *cfg)
3780 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3782 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3787 /* Port mirroring. */
3789 static struct mirror *
3790 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3794 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3795 if (uuid_equals(uuid, &m->uuid)) {
3803 bridge_configure_mirrors(struct bridge *br)
3805 const struct ovsdb_datum *mc;
3806 unsigned long *flood_vlans;
3807 struct mirror *m, *next;
3810 /* Get rid of deleted mirrors. */
3811 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3812 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3813 union ovsdb_atom atom;
3815 atom.uuid = m->uuid;
3816 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3821 /* Add new mirrors and reconfigure existing ones. */
3822 for (i = 0; i < br->cfg->n_mirrors; i++) {
3823 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3824 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3826 m = mirror_create(br, cfg);
3829 if (!mirror_configure(m)) {
3834 /* Update flooded vlans (for RSPAN). */
3835 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3836 br->cfg->n_flood_vlans);
3837 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3838 bitmap_free(flood_vlans);
3841 static struct mirror *
3842 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3846 m = xzalloc(sizeof *m);
3847 m->uuid = cfg->header_.uuid;
3848 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3850 m->name = xstrdup(cfg->name);
3856 mirror_destroy(struct mirror *m)
3859 struct bridge *br = m->bridge;
3862 ofproto_mirror_unregister(br->ofproto, m);
3865 hmap_remove(&br->mirrors, &m->hmap_node);
3872 mirror_collect_ports(struct mirror *m,
3873 struct ovsrec_port **in_ports, int n_in_ports,
3874 void ***out_portsp, size_t *n_out_portsp)
3876 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3877 size_t n_out_ports = 0;
3880 for (i = 0; i < n_in_ports; i++) {
3881 const char *name = in_ports[i]->name;
3882 struct port *port = port_lookup(m->bridge, name);
3884 out_ports[n_out_ports++] = port;
3886 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3887 "port %s", m->bridge->name, m->name, name);
3890 *out_portsp = out_ports;
3891 *n_out_portsp = n_out_ports;
3895 mirror_configure(struct mirror *m)
3897 const struct ovsrec_mirror *cfg = m->cfg;
3898 struct ofproto_mirror_settings s;
3901 if (strcmp(cfg->name, m->name)) {
3903 m->name = xstrdup(cfg->name);
3907 /* Get output port or VLAN. */
3908 if (cfg->output_port) {
3909 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3910 if (!s.out_bundle) {
3911 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3912 m->bridge->name, m->name);
3915 s.out_vlan = UINT16_MAX;
3917 if (cfg->output_vlan) {
3918 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3919 "output vlan; ignoring output vlan",
3920 m->bridge->name, m->name);
3922 } else if (cfg->output_vlan) {
3923 /* The database should prevent invalid VLAN values. */
3924 s.out_bundle = NULL;
3925 s.out_vlan = *cfg->output_vlan;
3927 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3928 m->bridge->name, m->name);
3932 /* Get port selection. */
3933 if (cfg->select_all) {
3934 size_t n_ports = hmap_count(&m->bridge->ports);
3935 void **ports = xmalloc(n_ports * sizeof *ports);
3940 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3950 /* Get ports, dropping ports that don't exist.
3951 * The IDL ensures that there are no duplicates. */
3952 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3953 &s.srcs, &s.n_srcs);
3954 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3955 &s.dsts, &s.n_dsts);
3958 /* Get VLAN selection. */
3959 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3962 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3965 if (s.srcs != s.dsts) {
3974 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3976 * This is deprecated. It is only for compatibility with broken device drivers
3977 * in old versions of Linux that do not properly support VLANs when VLAN
3978 * devices are not used. When broken device drivers are no longer in
3979 * widespread use, we will delete these interfaces. */
3981 static struct ovsrec_port **recs;
3982 static size_t n_recs, allocated_recs;
3984 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3985 * splinters are reconfigured. */
3987 register_rec(struct ovsrec_port *rec)
3989 if (n_recs >= allocated_recs) {
3990 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3992 recs[n_recs++] = rec;
3995 /* Frees all of the ports registered with register_reg(). */
3997 free_registered_recs(void)
4001 for (i = 0; i < n_recs; i++) {
4002 struct ovsrec_port *port = recs[i];
4005 for (j = 0; j < port->n_interfaces; j++) {
4006 struct ovsrec_interface *iface = port->interfaces[j];
4011 smap_destroy(&port->other_config);
4012 free(port->interfaces);
4020 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4023 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4025 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4029 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4032 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4033 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4034 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4037 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4038 * use, returns NULL.
4040 * Updates 'vlan_splinters_enabled_anywhere'. */
4041 static unsigned long int *
4042 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4044 unsigned long int *splinter_vlans;
4045 struct sset splinter_ifaces;
4046 const char *real_dev_name;
4047 struct shash *real_devs;
4048 struct shash_node *node;
4052 /* Free space allocated for synthesized ports and interfaces, since we're
4053 * in the process of reconstructing all of them. */
4054 free_registered_recs();
4056 splinter_vlans = bitmap_allocate(4096);
4057 sset_init(&splinter_ifaces);
4058 vlan_splinters_enabled_anywhere = false;
4059 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4060 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4063 for (j = 0; j < br_cfg->n_ports; j++) {
4064 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4067 for (k = 0; k < port_cfg->n_interfaces; k++) {
4068 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4070 if (vlan_splinters_is_enabled(iface_cfg)) {
4071 vlan_splinters_enabled_anywhere = true;
4072 sset_add(&splinter_ifaces, iface_cfg->name);
4073 vlan_bitmap_from_array__(port_cfg->trunks,
4079 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4080 bitmap_set1(splinter_vlans, *port_cfg->tag);
4085 if (!vlan_splinters_enabled_anywhere) {
4086 free(splinter_vlans);
4087 sset_destroy(&splinter_ifaces);
4091 HMAP_FOR_EACH (br, node, &all_bridges) {
4093 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4097 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4098 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4099 * device to be created for it. */
4100 bitmap_set0(splinter_vlans, 0);
4101 bitmap_set0(splinter_vlans, 4095);
4103 /* Delete all VLAN devices that we don't need. */
4105 real_devs = vlandev_get_real_devs();
4106 SHASH_FOR_EACH (node, real_devs) {
4107 const struct vlan_real_dev *real_dev = node->data;
4108 const struct vlan_dev *vlan_dev;
4109 bool real_dev_has_splinters;
4111 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4113 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4114 if (!real_dev_has_splinters
4115 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4116 struct netdev *netdev;
4118 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4119 if (!netdev_get_in4(netdev, NULL, NULL) ||
4120 !netdev_get_in6(netdev, NULL)) {
4121 /* It has an IP address configured, so we don't own
4122 * it. Don't delete it. */
4124 vlandev_del(vlan_dev->name);
4126 netdev_close(netdev);
4133 /* Add all VLAN devices that we need. */
4134 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4137 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4138 if (!vlandev_get_name(real_dev_name, vid)) {
4139 vlandev_add(real_dev_name, vid);
4146 sset_destroy(&splinter_ifaces);
4148 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4149 free(splinter_vlans);
4152 return splinter_vlans;
4155 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4158 configure_splinter_port(struct port *port)
4160 struct ofproto *ofproto = port->bridge->ofproto;
4161 ofp_port_t realdev_ofp_port;
4162 const char *realdev_name;
4163 struct iface *vlandev, *realdev;
4165 ofproto_bundle_unregister(port->bridge->ofproto, port);
4167 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4170 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4171 realdev = iface_lookup(port->bridge, realdev_name);
4172 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4174 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4178 static struct ovsrec_port *
4179 synthesize_splinter_port(const char *real_dev_name,
4180 const char *vlan_dev_name, int vid)
4182 struct ovsrec_interface *iface;
4183 struct ovsrec_port *port;
4185 iface = xmalloc(sizeof *iface);
4186 ovsrec_interface_init(iface);
4187 iface->name = xstrdup(vlan_dev_name);
4188 iface->type = "system";
4190 port = xmalloc(sizeof *port);
4191 ovsrec_port_init(port);
4192 port->interfaces = xmemdup(&iface, sizeof iface);
4193 port->n_interfaces = 1;
4194 port->name = xstrdup(vlan_dev_name);
4195 port->vlan_mode = "splinter";
4196 port->tag = xmalloc(sizeof *port->tag);
4199 smap_add(&port->other_config, "realdev", real_dev_name);
4205 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4206 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4207 * 1-bit in the 'splinter_vlans' bitmap. */
4209 add_vlan_splinter_ports(struct bridge *br,
4210 const unsigned long int *splinter_vlans,
4211 struct shash *ports)
4215 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4216 * we're modifying 'ports'. */
4217 for (i = 0; i < br->cfg->n_ports; i++) {
4218 const char *name = br->cfg->ports[i]->name;
4219 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4222 for (j = 0; j < port_cfg->n_interfaces; j++) {
4223 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4225 if (vlan_splinters_is_enabled(iface_cfg)) {
4226 const char *real_dev_name;
4229 real_dev_name = iface_cfg->name;
4230 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4231 const char *vlan_dev_name;
4233 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4235 && !shash_find(ports, vlan_dev_name)) {
4236 shash_add(ports, vlan_dev_name,
4237 synthesize_splinter_port(
4238 real_dev_name, vlan_dev_name, vid));
4247 mirror_refresh_stats(struct mirror *m)
4249 struct ofproto *ofproto = m->bridge->ofproto;
4250 uint64_t tx_packets, tx_bytes;
4253 size_t stat_cnt = 0;
4255 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4256 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4260 if (tx_packets != UINT64_MAX) {
4261 keys[stat_cnt] = "tx_packets";
4262 values[stat_cnt] = tx_packets;
4265 if (tx_bytes != UINT64_MAX) {
4266 keys[stat_cnt] = "tx_bytes";
4267 values[stat_cnt] = tx_bytes;
4271 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);