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 Flow_Sample_Collector_Set row is valid. */
989 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
990 const struct bridge *br)
992 return fscs->ipfix && fscs->bridge == br->cfg;
995 /* Set IPFIX configuration on 'br'. */
997 bridge_configure_ipfix(struct bridge *br)
999 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1000 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1001 struct ofproto_ipfix_bridge_exporter_options be_opts;
1002 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1003 size_t n_fe_opts = 0;
1005 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1006 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1011 if (!be_cfg && n_fe_opts == 0) {
1012 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1017 memset(&be_opts, 0, sizeof be_opts);
1019 sset_init(&be_opts.targets);
1020 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1022 if (be_cfg->sampling) {
1023 be_opts.sampling_rate = *be_cfg->sampling;
1025 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1027 if (be_cfg->obs_domain_id) {
1028 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1030 if (be_cfg->obs_point_id) {
1031 be_opts.obs_point_id = *be_cfg->obs_point_id;
1033 if (be_cfg->cache_active_timeout) {
1034 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1036 if (be_cfg->cache_max_flows) {
1037 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1041 if (n_fe_opts > 0) {
1042 struct ofproto_ipfix_flow_exporter_options *opts;
1043 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1045 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1046 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1047 opts->collector_set_id = fe_cfg->id;
1048 sset_init(&opts->targets);
1049 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1050 fe_cfg->ipfix->n_targets);
1051 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1052 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1053 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1054 ? *fe_cfg->ipfix->cache_max_flows : 0;
1060 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1064 sset_destroy(&be_opts.targets);
1067 if (n_fe_opts > 0) {
1068 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1070 for (i = 0; i < n_fe_opts; i++) {
1071 sset_destroy(&opts->targets);
1079 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1080 struct ofproto_port_stp_settings *port_s,
1081 int *port_num_counter, unsigned long *port_num_bitmap)
1083 const char *config_str;
1084 struct iface *iface;
1086 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1087 port_s->enable = false;
1090 port_s->enable = true;
1093 /* STP over bonds is not supported. */
1094 if (!list_is_singleton(&port->ifaces)) {
1095 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1097 port_s->enable = false;
1101 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1103 /* Internal ports shouldn't participate in spanning tree, so
1105 if (!strcmp(iface->type, "internal")) {
1106 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1107 port_s->enable = false;
1111 /* STP on mirror output ports is not supported. */
1112 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1113 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1114 port_s->enable = false;
1118 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1120 unsigned long int port_num = strtoul(config_str, NULL, 0);
1121 int port_idx = port_num - 1;
1123 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1124 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1125 port_s->enable = false;
1129 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1130 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1131 port->name, port_num);
1132 port_s->enable = false;
1135 bitmap_set1(port_num_bitmap, port_idx);
1136 port_s->port_num = port_idx;
1138 if (*port_num_counter >= STP_MAX_PORTS) {
1139 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1140 port_s->enable = false;
1144 port_s->port_num = (*port_num_counter)++;
1147 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1149 port_s->path_cost = strtoul(config_str, NULL, 10);
1151 enum netdev_features current;
1154 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1155 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1156 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1159 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1161 port_s->priority = strtoul(config_str, NULL, 0);
1163 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1167 /* Set spanning tree configuration on 'br'. */
1169 bridge_configure_stp(struct bridge *br)
1171 if (!br->cfg->stp_enable) {
1172 ofproto_set_stp(br->ofproto, NULL);
1174 struct ofproto_stp_settings br_s;
1175 const char *config_str;
1177 int port_num_counter;
1178 unsigned long *port_num_bitmap;
1180 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1182 uint8_t ea[ETH_ADDR_LEN];
1184 if (eth_addr_from_string(config_str, ea)) {
1185 br_s.system_id = eth_addr_to_uint64(ea);
1187 br_s.system_id = eth_addr_to_uint64(br->ea);
1188 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1189 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1192 br_s.system_id = eth_addr_to_uint64(br->ea);
1195 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1197 br_s.priority = strtoul(config_str, NULL, 0);
1199 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1202 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1204 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1206 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1209 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1211 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1213 br_s.max_age = STP_DEFAULT_MAX_AGE;
1216 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1218 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1220 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1223 /* Configure STP on the bridge. */
1224 if (ofproto_set_stp(br->ofproto, &br_s)) {
1225 VLOG_ERR("bridge %s: could not enable STP", br->name);
1229 /* Users must either set the port number with the "stp-port-num"
1230 * configuration on all ports or none. If manual configuration
1231 * is not done, then we allocate them sequentially. */
1232 port_num_counter = 0;
1233 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1234 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1235 struct ofproto_port_stp_settings port_s;
1236 struct iface *iface;
1238 port_configure_stp(br->ofproto, port, &port_s,
1239 &port_num_counter, port_num_bitmap);
1241 /* As bonds are not supported, just apply configuration to
1242 * all interfaces. */
1243 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1244 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1246 VLOG_ERR("port %s: could not enable STP", port->name);
1252 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1253 && port_num_counter) {
1254 VLOG_ERR("bridge %s: must manually configure all STP port "
1255 "IDs or none, disabling", br->name);
1256 ofproto_set_stp(br->ofproto, NULL);
1258 bitmap_free(port_num_bitmap);
1263 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1265 const struct port *port = port_lookup(br, name);
1266 return port && port_is_bond_fake_iface(port);
1270 port_is_bond_fake_iface(const struct port *port)
1272 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1276 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1278 struct bridge *br, *next;
1279 struct shash new_br;
1282 /* Collect new bridges' names and types. */
1283 shash_init(&new_br);
1284 for (i = 0; i < cfg->n_bridges; i++) {
1285 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1286 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1288 if (strchr(br_cfg->name, '/')) {
1289 /* Prevent remote ovsdb-server users from accessing arbitrary
1290 * directories, e.g. consider a bridge named "../../../etc/". */
1291 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1293 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1294 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1298 /* Get rid of deleted bridges or those whose types have changed.
1299 * Update 'cfg' of bridges that still exist. */
1300 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1301 br->cfg = shash_find_data(&new_br, br->name);
1302 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1303 br->cfg->datapath_type))) {
1308 /* Add new bridges. */
1309 for (i = 0; i < cfg->n_bridges; i++) {
1310 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1311 struct bridge *br = bridge_lookup(br_cfg->name);
1313 bridge_create(br_cfg);
1317 shash_destroy(&new_br);
1321 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1323 struct bridge *br = iface->port->bridge;
1325 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1326 iface->ofp_port = ofp_port;
1327 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1328 hash_ofp_port(ofp_port));
1329 iface_set_ofport(iface->cfg, ofp_port);
1332 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1333 * Returns 0 if successful, otherwise a positive errno value. */
1335 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1336 struct netdev *netdev)
1340 error = netdev_set_config(netdev, &iface_cfg->options);
1342 VLOG_WARN("could not configure network device %s (%s)",
1343 iface_cfg->name, ovs_strerror(error));
1348 /* This function determines whether 'ofproto_port', which is attached to
1349 * br->ofproto's datapath, is one that we want in 'br'.
1351 * If it is, it returns true, after creating an iface (if necessary),
1352 * configuring the iface's netdev according to the iface's options, and setting
1353 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1355 * If, on the other hand, 'port' should be removed, it returns false. The
1356 * caller should later detach the port from br->ofproto. */
1358 bridge_refresh_one_ofp_port(struct bridge *br,
1359 const struct ofproto_port *ofproto_port)
1361 const char *name = ofproto_port->name;
1362 const char *type = ofproto_port->type;
1363 ofp_port_t ofp_port = ofproto_port->ofp_port;
1365 struct iface *iface = iface_lookup(br, name);
1367 /* Check that the name-to-number mapping is one-to-one. */
1368 if (iface->ofp_port != OFPP_NONE) {
1369 VLOG_WARN("bridge %s: interface %s reported twice",
1372 } else if (iface_from_ofp_port(br, ofp_port)) {
1373 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1374 br->name, ofp_port);
1378 /* There's a configured interface named 'name'. */
1379 if (strcmp(type, iface->type)
1380 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1381 /* It's the wrong type, or it's the right type but can't be
1382 * configured as the user requested, so we must destroy it. */
1385 /* It's the right type and configured correctly. Keep it. */
1386 iface_set_ofp_port(iface, ofp_port);
1389 } else if (bridge_has_bond_fake_iface(br, name)
1390 && !strcmp(type, "internal")) {
1391 /* It's a bond fake iface. Keep it. */
1394 /* There's no configured interface named 'name', but there might be an
1395 * interface of that name queued to be created.
1397 * If there is, and it has the correct type, then try to configure it
1398 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1399 * delete it and later try to re-add it. */
1400 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1402 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1403 && iface_create(br, if_cfg, ofp_port));
1407 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1408 * consistent with the ofp_ports in "br->ofproto". */
1410 bridge_refresh_ofp_port(struct bridge *br)
1412 struct ofproto_port_dump dump;
1413 struct ofproto_port ofproto_port;
1414 struct port *port, *port_next;
1416 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1417 hmap_clear(&br->ifaces);
1418 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1419 struct iface *iface;
1421 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1422 iface->ofp_port = OFPP_NONE;
1426 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1427 * already exist in the datapath and promote them to full fledged "struct
1428 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1429 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1430 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1431 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1432 garbage->ofp_port = ofproto_port.ofp_port;
1433 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1437 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1438 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1439 * later they can be added to ofproto. */
1440 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1441 struct iface *iface, *iface_next;
1443 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1444 if (iface->ofp_port == OFPP_NONE) {
1445 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1446 iface_destroy(iface);
1450 if (list_is_empty(&port->ifaces)) {
1456 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1457 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1458 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1461 * If successful, returns 0 and stores the network device in '*netdevp'. On
1462 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1464 iface_do_create(const struct bridge *br,
1465 const struct if_cfg *if_cfg,
1466 ofp_port_t *ofp_portp, struct netdev **netdevp)
1468 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1469 const struct ovsrec_port *port_cfg = if_cfg->parent;
1470 struct netdev *netdev = NULL;
1473 if (netdev_is_reserved_name(iface_cfg->name)) {
1474 VLOG_WARN("could not create interface %s, name is reserved",
1480 error = netdev_open(iface_cfg->name,
1481 iface_get_type(iface_cfg, br->cfg), &netdev);
1483 VLOG_WARN("could not open network device %s (%s)",
1484 iface_cfg->name, ovs_strerror(error));
1488 error = iface_set_netdev_config(iface_cfg, netdev);
1493 if (*ofp_portp == OFPP_NONE) {
1494 ofp_port_t ofp_port = if_cfg->ofport;
1496 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1500 *ofp_portp = ofp_port;
1502 VLOG_INFO("bridge %s: added interface %s on port %d",
1503 br->name, iface_cfg->name, *ofp_portp);
1505 VLOG_DBG("bridge %s: interface %s is on port %d",
1506 br->name, iface_cfg->name, *ofp_portp);
1509 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1510 || iface_is_internal(iface_cfg, br->cfg)) {
1511 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1519 netdev_close(netdev);
1523 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1524 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1525 * automatically allocated for the iface. Takes ownership of and
1526 * deallocates 'if_cfg'.
1528 * Return true if an iface is successfully created, false otherwise. */
1530 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1532 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1533 const struct ovsrec_port *port_cfg = if_cfg->parent;
1535 struct netdev *netdev;
1536 struct iface *iface;
1541 /* Do the bits that can fail up front.
1543 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1544 * internal datastructures may not be consistent. Eventually, when port
1545 * additions and deletions are cheaper, these calls should be removed. */
1547 ovs_assert(!iface_lookup(br, iface_cfg->name));
1548 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1551 iface_set_ofport(iface_cfg, OFPP_NONE);
1552 iface_clear_db_record(iface_cfg);
1557 /* Get or create the port structure. */
1558 port = port_lookup(br, port_cfg->name);
1560 port = port_create(br, port_cfg);
1563 /* Create the iface structure. */
1564 iface = xzalloc(sizeof *iface);
1565 list_push_back(&port->ifaces, &iface->port_elem);
1566 hmap_insert(&br->iface_by_name, &iface->name_node,
1567 hash_string(iface_cfg->name, 0));
1569 iface->name = xstrdup(iface_cfg->name);
1570 iface->ofp_port = OFPP_NONE;
1571 iface->netdev = netdev;
1572 iface->type = iface_get_type(iface_cfg, br->cfg);
1573 iface->cfg = iface_cfg;
1575 iface_set_ofp_port(iface, ofp_port);
1577 /* Populate initial status in database. */
1578 iface_refresh_stats(iface);
1579 iface_refresh_status(iface);
1581 /* Add bond fake iface if necessary. */
1582 if (port_is_bond_fake_iface(port)) {
1583 struct ofproto_port ofproto_port;
1585 if (ofproto_port_query_by_name(br->ofproto, port->name,
1587 struct netdev *netdev;
1590 error = netdev_open(port->name, "internal", &netdev);
1592 ofp_port_t fake_ofp_port = if_cfg->ofport;
1594 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1595 netdev_close(netdev);
1597 VLOG_WARN("could not open network device %s (%s)",
1598 port->name, ovs_strerror(error));
1601 /* Already exists, nothing to do. */
1602 ofproto_port_destroy(&ofproto_port);
1607 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1613 /* Set forward BPDU option. */
1615 bridge_configure_forward_bpdu(struct bridge *br)
1617 ofproto_set_forward_bpdu(br->ofproto,
1618 smap_get_bool(&br->cfg->other_config,
1623 /* Set MAC learning table configuration for 'br'. */
1625 bridge_configure_mac_table(struct bridge *br)
1627 const char *idle_time_str;
1630 const char *mac_table_size_str;
1633 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1634 idle_time = (idle_time_str && atoi(idle_time_str)
1635 ? atoi(idle_time_str)
1636 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1638 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1639 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1640 ? atoi(mac_table_size_str)
1643 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1647 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1648 struct iface **hw_addr_iface)
1650 struct hmapx mirror_output_ports;
1653 bool found_addr = false;
1657 *hw_addr_iface = NULL;
1659 /* Did the user request a particular MAC? */
1660 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1661 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1662 if (eth_addr_is_multicast(ea)) {
1663 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1664 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1665 } else if (eth_addr_is_zero(ea)) {
1666 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1672 /* Mirror output ports don't participate in picking the local hardware
1673 * address. ofproto can't help us find out whether a given port is a
1674 * mirror output because we haven't configured mirrors yet, so we need to
1675 * accumulate them ourselves. */
1676 hmapx_init(&mirror_output_ports);
1677 for (i = 0; i < br->cfg->n_mirrors; i++) {
1678 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1679 if (m->output_port) {
1680 hmapx_add(&mirror_output_ports, m->output_port);
1684 /* Otherwise choose the minimum non-local MAC address among all of the
1686 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1687 uint8_t iface_ea[ETH_ADDR_LEN];
1688 struct iface *candidate;
1689 struct iface *iface;
1691 /* Mirror output ports don't participate. */
1692 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1696 /* Choose the MAC address to represent the port. */
1698 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1699 /* Find the interface with this Ethernet address (if any) so that
1700 * we can provide the correct devname to the caller. */
1701 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1702 uint8_t candidate_ea[ETH_ADDR_LEN];
1703 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1704 && eth_addr_equals(iface_ea, candidate_ea)) {
1709 /* Choose the interface whose MAC address will represent the port.
1710 * The Linux kernel bonding code always chooses the MAC address of
1711 * the first slave added to a bond, and the Fedora networking
1712 * scripts always add slaves to a bond in alphabetical order, so
1713 * for compatibility we choose the interface with the name that is
1714 * first in alphabetical order. */
1715 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1716 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1721 /* The local port doesn't count (since we're trying to choose its
1722 * MAC address anyway). */
1723 if (iface->ofp_port == OFPP_LOCAL) {
1728 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1734 /* Compare against our current choice. */
1735 if (!eth_addr_is_multicast(iface_ea) &&
1736 !eth_addr_is_local(iface_ea) &&
1737 !eth_addr_is_reserved(iface_ea) &&
1738 !eth_addr_is_zero(iface_ea) &&
1739 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1741 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1742 *hw_addr_iface = iface;
1748 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1749 *hw_addr_iface = NULL;
1752 hmapx_destroy(&mirror_output_ports);
1755 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1756 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1757 * an interface on 'br', then that interface must be passed in as
1758 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1759 * 'hw_addr_iface' must be passed in as a null pointer. */
1761 bridge_pick_datapath_id(struct bridge *br,
1762 const uint8_t bridge_ea[ETH_ADDR_LEN],
1763 struct iface *hw_addr_iface)
1766 * The procedure for choosing a bridge MAC address will, in the most
1767 * ordinary case, also choose a unique MAC that we can use as a datapath
1768 * ID. In some special cases, though, multiple bridges will end up with
1769 * the same MAC address. This is OK for the bridges, but it will confuse
1770 * the OpenFlow controller, because each datapath needs a unique datapath
1773 * Datapath IDs must be unique. It is also very desirable that they be
1774 * stable from one run to the next, so that policy set on a datapath
1777 const char *datapath_id;
1780 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1781 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1785 if (!hw_addr_iface) {
1787 * A purely internal bridge, that is, one that has no non-virtual
1788 * network devices on it at all, is difficult because it has no
1789 * natural unique identifier at all.
1791 * When the host is a XenServer, we handle this case by hashing the
1792 * host's UUID with the name of the bridge. Names of bridges are
1793 * persistent across XenServer reboots, although they can be reused if
1794 * an internal network is destroyed and then a new one is later
1795 * created, so this is fairly effective.
1797 * When the host is not a XenServer, we punt by using a random MAC
1798 * address on each run.
1800 const char *host_uuid = xenserver_get_host_uuid();
1802 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1803 dpid = dpid_from_hash(combined, strlen(combined));
1809 return eth_addr_to_uint64(bridge_ea);
1813 dpid_from_hash(const void *data, size_t n)
1815 uint8_t hash[SHA1_DIGEST_SIZE];
1817 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1818 sha1_bytes(data, n, hash);
1819 eth_addr_mark_random(hash);
1820 return eth_addr_to_uint64(hash);
1824 iface_refresh_status(struct iface *iface)
1828 enum netdev_features current;
1832 uint8_t mac[ETH_ADDR_LEN];
1836 if (iface_is_synthetic(iface)) {
1842 if (!netdev_get_status(iface->netdev, &smap)) {
1843 ovsrec_interface_set_status(iface->cfg, &smap);
1845 ovsrec_interface_set_status(iface->cfg, NULL);
1848 smap_destroy(&smap);
1850 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1851 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1853 ovsrec_interface_set_duplex(iface->cfg,
1854 netdev_features_is_full_duplex(current)
1856 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1858 ovsrec_interface_set_duplex(iface->cfg, NULL);
1859 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1862 error = netdev_get_mtu(iface->netdev, &mtu);
1865 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1867 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1870 error = netdev_get_etheraddr(iface->netdev, mac);
1872 char mac_string[32];
1874 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1875 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1877 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1880 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1881 * if there is no valid ifindex number. */
1882 ifindex64 = netdev_get_ifindex(iface->netdev);
1883 if (ifindex64 < 0) {
1886 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1889 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1892 iface_refresh_cfm_stats(struct iface *iface)
1894 const struct ovsrec_interface *cfg = iface->cfg;
1895 struct ofproto_cfm_status status;
1897 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1898 iface->ofp_port, &status)) {
1899 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1900 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1901 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1902 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1903 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1904 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1906 const char *reasons[CFM_FAULT_N_REASONS];
1907 int64_t cfm_health = status.health;
1908 int64_t cfm_flap_count = status.flap_count;
1909 bool faulted = status.faults != 0;
1912 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1915 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1916 int reason = 1 << i;
1917 if (status.faults & reason) {
1918 reasons[j++] = cfm_fault_reason_to_str(reason);
1921 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1923 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1925 if (status.remote_opstate >= 0) {
1926 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1927 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1929 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1932 ovsrec_interface_set_cfm_remote_mpids(cfg,
1933 (const int64_t *)status.rmps,
1935 if (cfm_health >= 0) {
1936 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1938 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1946 iface_refresh_stats(struct iface *iface)
1948 #define IFACE_STATS \
1949 IFACE_STAT(rx_packets, "rx_packets") \
1950 IFACE_STAT(tx_packets, "tx_packets") \
1951 IFACE_STAT(rx_bytes, "rx_bytes") \
1952 IFACE_STAT(tx_bytes, "tx_bytes") \
1953 IFACE_STAT(rx_dropped, "rx_dropped") \
1954 IFACE_STAT(tx_dropped, "tx_dropped") \
1955 IFACE_STAT(rx_errors, "rx_errors") \
1956 IFACE_STAT(tx_errors, "tx_errors") \
1957 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1958 IFACE_STAT(rx_over_errors, "rx_over_err") \
1959 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1960 IFACE_STAT(collisions, "collisions")
1962 #define IFACE_STAT(MEMBER, NAME) + 1
1963 enum { N_IFACE_STATS = IFACE_STATS };
1965 int64_t values[N_IFACE_STATS];
1966 char *keys[N_IFACE_STATS];
1969 struct netdev_stats stats;
1971 if (iface_is_synthetic(iface)) {
1975 /* Intentionally ignore return value, since errors will set 'stats' to
1976 * all-1s, and we will deal with that correctly below. */
1977 netdev_get_stats(iface->netdev, &stats);
1979 /* Copy statistics into keys[] and values[]. */
1981 #define IFACE_STAT(MEMBER, NAME) \
1982 if (stats.MEMBER != UINT64_MAX) { \
1984 values[n] = stats.MEMBER; \
1989 ovs_assert(n <= N_IFACE_STATS);
1991 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1996 br_refresh_stp_status(struct bridge *br)
1998 struct smap smap = SMAP_INITIALIZER(&smap);
1999 struct ofproto *ofproto = br->ofproto;
2000 struct ofproto_stp_status status;
2002 if (ofproto_get_stp_status(ofproto, &status)) {
2006 if (!status.enabled) {
2007 ovsrec_bridge_set_status(br->cfg, NULL);
2011 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2012 STP_ID_ARGS(status.bridge_id));
2013 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2014 STP_ID_ARGS(status.designated_root));
2015 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2017 ovsrec_bridge_set_status(br->cfg, &smap);
2018 smap_destroy(&smap);
2022 port_refresh_stp_status(struct port *port)
2024 struct ofproto *ofproto = port->bridge->ofproto;
2025 struct iface *iface;
2026 struct ofproto_port_stp_status status;
2028 int64_t int_values[3];
2031 if (port_is_synthetic(port)) {
2035 /* STP doesn't currently support bonds. */
2036 if (!list_is_singleton(&port->ifaces)) {
2037 ovsrec_port_set_status(port->cfg, NULL);
2041 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2043 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2047 if (!status.enabled) {
2048 ovsrec_port_set_status(port->cfg, NULL);
2049 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2053 /* Set Status column. */
2055 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2056 smap_add(&smap, "stp_state", stp_state_name(status.state));
2057 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2058 smap_add(&smap, "stp_role", stp_role_name(status.role));
2059 ovsrec_port_set_status(port->cfg, &smap);
2060 smap_destroy(&smap);
2062 /* Set Statistics column. */
2063 keys[0] = "stp_tx_count";
2064 int_values[0] = status.tx_count;
2065 keys[1] = "stp_rx_count";
2066 int_values[1] = status.rx_count;
2067 keys[2] = "stp_error_count";
2068 int_values[2] = status.error_count;
2070 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2071 ARRAY_SIZE(int_values));
2075 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2077 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2081 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2083 bool enable = enable_system_stats(cfg);
2085 system_stats_enable(enable);
2087 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2092 run_system_stats(void)
2094 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2097 stats = system_stats_run();
2099 struct ovsdb_idl_txn *txn;
2100 struct ovsdb_datum datum;
2102 txn = ovsdb_idl_txn_create(idl);
2103 ovsdb_datum_from_smap(&datum, stats);
2104 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2106 ovsdb_idl_txn_commit(txn);
2107 ovsdb_idl_txn_destroy(txn);
2113 static inline const char *
2114 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2117 case OFPCR12_ROLE_EQUAL:
2119 case OFPCR12_ROLE_MASTER:
2121 case OFPCR12_ROLE_SLAVE:
2123 case OFPCR12_ROLE_NOCHANGE:
2125 return "*** INVALID ROLE ***";
2130 refresh_controller_status(void)
2134 const struct ovsrec_controller *cfg;
2138 /* Accumulate status for controllers on all bridges. */
2139 HMAP_FOR_EACH (br, node, &all_bridges) {
2140 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2143 /* Update each controller in the database with current status. */
2144 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2145 struct ofproto_controller_info *cinfo =
2146 shash_find_data(&info, cfg->target);
2149 struct smap smap = SMAP_INITIALIZER(&smap);
2150 const char **values = cinfo->pairs.values;
2151 const char **keys = cinfo->pairs.keys;
2154 for (i = 0; i < cinfo->pairs.n; i++) {
2155 smap_add(&smap, keys[i], values[i]);
2158 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2159 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2161 ovsrec_controller_set_status(cfg, &smap);
2162 smap_destroy(&smap);
2164 ovsrec_controller_set_is_connected(cfg, false);
2165 ovsrec_controller_set_role(cfg, NULL);
2166 ovsrec_controller_set_status(cfg, NULL);
2170 ofproto_free_ofproto_controller_info(&info);
2175 * Some information in the database must be kept as up-to-date as possible to
2176 * allow controllers to respond rapidly to network outages. We call these
2177 * statistics "instant" stats.
2179 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2180 * assuming that they've changed. The only means we have to determine whether
2181 * they have changed are:
2183 * - Try to commit changes to the database. If nothing changed, then
2184 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2187 * - instant_stats_run() is called late in the run loop, after anything that
2188 * might change any of the instant stats.
2190 * We use these two facts together to avoid waking the process up every
2191 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2194 /* Minimum interval between writing updates to the instant stats to the
2196 #define INSTANT_INTERVAL_MSEC 100
2198 /* Current instant stats database transaction, NULL if there is no ongoing
2200 static struct ovsdb_idl_txn *instant_txn;
2202 /* Next time (in msec on monotonic clock) at which we will update the instant
2204 static long long int instant_next_txn = LLONG_MIN;
2206 /* True if the run loop has run since we last saw that the instant stats were
2207 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2208 * to refresh the instant stats. */
2209 static bool instant_stats_could_have_changed;
2212 instant_stats_run(void)
2214 enum ovsdb_idl_txn_status status;
2216 instant_stats_could_have_changed = true;
2221 if (time_msec() < instant_next_txn) {
2224 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2226 instant_txn = ovsdb_idl_txn_create(idl);
2227 HMAP_FOR_EACH (br, node, &all_bridges) {
2228 struct iface *iface;
2231 br_refresh_stp_status(br);
2233 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2234 port_refresh_stp_status(port);
2237 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2238 enum netdev_flags flags;
2240 const char *link_state;
2241 int64_t link_resets;
2244 if (iface_is_synthetic(iface)) {
2248 current = ofproto_port_is_lacp_current(br->ofproto,
2252 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2254 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2257 error = netdev_get_flags(iface->netdev, &flags);
2259 const char *state = flags & NETDEV_UP ? "up" : "down";
2260 ovsrec_interface_set_admin_state(iface->cfg, state);
2262 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2265 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2266 ovsrec_interface_set_link_state(iface->cfg, link_state);
2268 link_resets = netdev_get_carrier_resets(iface->netdev);
2269 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2271 iface_refresh_cfm_stats(iface);
2274 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2276 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2277 smap_destroy(&smap);
2282 status = ovsdb_idl_txn_commit(instant_txn);
2283 if (status != TXN_INCOMPLETE) {
2284 ovsdb_idl_txn_destroy(instant_txn);
2287 if (status == TXN_UNCHANGED) {
2288 instant_stats_could_have_changed = false;
2293 instant_stats_wait(void)
2296 ovsdb_idl_txn_wait(instant_txn);
2297 } else if (instant_stats_could_have_changed) {
2298 poll_timer_wait_until(instant_next_txn);
2302 /* Performs periodic activity required by bridges that needs to be done with
2303 * the least possible latency.
2305 * It makes sense to call this function a couple of times per poll loop, to
2306 * provide a significant performance boost on some benchmarks with ofprotos
2307 * that use the ofproto-dpif implementation. */
2309 bridge_run_fast(void)
2316 ofproto_enumerate_types(&types);
2317 SSET_FOR_EACH (type, &types) {
2318 ofproto_type_run_fast(type);
2320 sset_destroy(&types);
2322 HMAP_FOR_EACH (br, node, &all_bridges) {
2323 ofproto_run_fast(br->ofproto);
2330 static struct ovsrec_open_vswitch null_cfg;
2331 const struct ovsrec_open_vswitch *cfg;
2332 struct ovsdb_idl_txn *reconf_txn = NULL;
2336 bool vlan_splinters_changed;
2339 ovsrec_open_vswitch_init(&null_cfg);
2341 /* (Re)configure if necessary. */
2342 if (!reconfiguring) {
2345 if (ovsdb_idl_is_lock_contended(idl)) {
2346 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2347 struct bridge *br, *next_br;
2349 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2350 "disabling this process (pid %ld) until it goes away",
2351 (long int) getpid());
2353 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2356 /* Since we will not be running system_stats_run() in this process
2357 * with the current situation of multiple ovs-vswitchd daemons,
2358 * disable system stats collection. */
2359 system_stats_enable(false);
2361 } else if (!ovsdb_idl_has_lock(idl)) {
2365 cfg = ovsrec_open_vswitch_first(idl);
2367 /* Initialize the ofproto library. This only needs to run once, but
2368 * it must be done after the configuration is set. If the
2369 * initialization has already occurred, bridge_init_ofproto()
2370 * returns immediately. */
2371 bridge_init_ofproto(cfg);
2373 /* Once the value of flow-restore-wait is false, we no longer should
2374 * check its value from the database. */
2375 if (cfg && ofproto_get_flow_restore_wait()) {
2376 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2377 "flow-restore-wait", false));
2380 /* Let each datapath type do the work that it needs to do. */
2382 ofproto_enumerate_types(&types);
2383 SSET_FOR_EACH (type, &types) {
2384 ofproto_type_run(type);
2386 sset_destroy(&types);
2388 /* Let each bridge do the work that it needs to do. */
2389 HMAP_FOR_EACH (br, node, &all_bridges) {
2390 ofproto_run(br->ofproto);
2393 /* Re-configure SSL. We do this on every trip through the main loop,
2394 * instead of just when the database changes, because the contents of the
2395 * key and certificate files can change without the database changing.
2397 * We do this before bridge_reconfigure() because that function might
2398 * initiate SSL connections and thus requires SSL to be configured. */
2399 if (cfg && cfg->ssl) {
2400 const struct ovsrec_ssl *ssl = cfg->ssl;
2402 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2403 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2406 if (!reconfiguring) {
2407 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2408 * usage has changed. */
2409 vlan_splinters_changed = false;
2410 if (vlan_splinters_enabled_anywhere) {
2411 HMAP_FOR_EACH (br, node, &all_bridges) {
2412 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2413 vlan_splinters_changed = true;
2419 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2420 idl_seqno = ovsdb_idl_get_seqno(idl);
2422 reconf_txn = ovsdb_idl_txn_create(idl);
2423 bridge_reconfigure(cfg);
2425 /* We still need to reconfigure to avoid dangling pointers to
2426 * now-destroyed ovsrec structures inside bridge data. */
2427 bridge_reconfigure(&null_cfg);
2432 if (reconfiguring) {
2434 reconf_txn = ovsdb_idl_txn_create(idl);
2437 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2438 reconfiguring = false;
2441 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2444 /* If we are completing our initial configuration for this run
2445 * of ovs-vswitchd, then keep the transaction around to monitor
2446 * it for completion. */
2447 if (!initial_config_done) {
2448 initial_config_done = true;
2449 daemonize_txn = reconf_txn;
2456 ovsdb_idl_txn_commit(reconf_txn);
2457 ovsdb_idl_txn_destroy(reconf_txn);
2461 if (daemonize_txn) {
2462 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2463 if (status != TXN_INCOMPLETE) {
2464 ovsdb_idl_txn_destroy(daemonize_txn);
2465 daemonize_txn = NULL;
2467 /* ovs-vswitchd has completed initialization, so allow the
2468 * process that forked us to exit successfully. */
2469 daemonize_complete();
2471 vlog_enable_async();
2473 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2477 /* Refresh interface and mirror stats if necessary. */
2478 if (time_msec() >= iface_stats_timer) {
2480 struct ovsdb_idl_txn *txn;
2482 txn = ovsdb_idl_txn_create(idl);
2483 HMAP_FOR_EACH (br, node, &all_bridges) {
2487 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2488 struct iface *iface;
2490 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2491 iface_refresh_stats(iface);
2492 iface_refresh_status(iface);
2496 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2497 mirror_refresh_stats(m);
2501 refresh_controller_status();
2502 ovsdb_idl_txn_commit(txn);
2503 ovsdb_idl_txn_destroy(txn); /* XXX */
2506 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2510 instant_stats_run();
2519 ovsdb_idl_wait(idl);
2520 if (daemonize_txn) {
2521 ovsdb_idl_txn_wait(daemonize_txn);
2524 if (reconfiguring) {
2525 poll_immediate_wake();
2529 ofproto_enumerate_types(&types);
2530 SSET_FOR_EACH (type, &types) {
2531 ofproto_type_wait(type);
2533 sset_destroy(&types);
2535 if (!hmap_is_empty(&all_bridges)) {
2538 HMAP_FOR_EACH (br, node, &all_bridges) {
2539 ofproto_wait(br->ofproto);
2541 poll_timer_wait_until(iface_stats_timer);
2544 system_stats_wait();
2545 instant_stats_wait();
2548 /* Adds some memory usage statistics for bridges into 'usage', for use with
2549 * memory_report(). */
2551 bridge_get_memory_usage(struct simap *usage)
2555 HMAP_FOR_EACH (br, node, &all_bridges) {
2556 ofproto_get_memory_usage(br->ofproto, usage);
2560 /* QoS unixctl user interface functions. */
2562 struct qos_unixctl_show_cbdata {
2564 struct iface *iface;
2568 qos_unixctl_show_queue(unsigned int queue_id,
2569 const struct smap *details,
2570 struct iface *iface,
2573 struct netdev_queue_stats stats;
2574 struct smap_node *node;
2577 ds_put_cstr(ds, "\n");
2579 ds_put_format(ds, "Queue %u:\n", queue_id);
2581 ds_put_cstr(ds, "Default:\n");
2584 SMAP_FOR_EACH (node, details) {
2585 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2588 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2590 if (stats.tx_packets != UINT64_MAX) {
2591 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2594 if (stats.tx_bytes != UINT64_MAX) {
2595 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2598 if (stats.tx_errors != UINT64_MAX) {
2599 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2602 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2603 queue_id, ovs_strerror(error));
2608 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2609 const char *argv[], void *aux OVS_UNUSED)
2611 struct ds ds = DS_EMPTY_INITIALIZER;
2612 struct smap smap = SMAP_INITIALIZER(&smap);
2613 struct iface *iface;
2615 struct smap_node *node;
2617 iface = iface_find(argv[1]);
2619 unixctl_command_reply_error(conn, "no such interface");
2623 netdev_get_qos(iface->netdev, &type, &smap);
2625 if (*type != '\0') {
2626 struct netdev_queue_dump dump;
2627 struct smap details;
2628 unsigned int queue_id;
2630 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2632 SMAP_FOR_EACH (node, &smap) {
2633 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2636 smap_init(&details);
2637 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2638 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2640 smap_destroy(&details);
2642 unixctl_command_reply(conn, ds_cstr(&ds));
2644 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2645 unixctl_command_reply_error(conn, ds_cstr(&ds));
2648 smap_destroy(&smap);
2652 /* Bridge reconfiguration functions. */
2654 bridge_create(const struct ovsrec_bridge *br_cfg)
2658 ovs_assert(!bridge_lookup(br_cfg->name));
2659 br = xzalloc(sizeof *br);
2661 br->name = xstrdup(br_cfg->name);
2662 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2665 /* Derive the default Ethernet address from the bridge's UUID. This should
2666 * be unique and it will be stable between ovs-vswitchd runs. */
2667 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2668 eth_addr_mark_random(br->default_ea);
2670 hmap_init(&br->ports);
2671 hmap_init(&br->ifaces);
2672 hmap_init(&br->iface_by_name);
2673 hmap_init(&br->mirrors);
2675 hmap_init(&br->if_cfg_todo);
2676 list_init(&br->ofpp_garbage);
2678 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2682 bridge_destroy(struct bridge *br)
2685 struct mirror *mirror, *next_mirror;
2686 struct port *port, *next_port;
2687 struct if_cfg *if_cfg, *next_if_cfg;
2688 struct ofpp_garbage *garbage, *next_garbage;
2690 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2693 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2694 mirror_destroy(mirror);
2696 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2697 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2700 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2701 &br->ofpp_garbage) {
2702 list_remove(&garbage->list_node);
2706 hmap_remove(&all_bridges, &br->node);
2707 ofproto_destroy(br->ofproto);
2708 hmap_destroy(&br->ifaces);
2709 hmap_destroy(&br->ports);
2710 hmap_destroy(&br->iface_by_name);
2711 hmap_destroy(&br->mirrors);
2712 hmap_destroy(&br->if_cfg_todo);
2719 static struct bridge *
2720 bridge_lookup(const char *name)
2724 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2725 if (!strcmp(br->name, name)) {
2732 /* Handle requests for a listing of all flows known by the OpenFlow
2733 * stack, including those normally hidden. */
2735 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2736 const char *argv[], void *aux OVS_UNUSED)
2741 br = bridge_lookup(argv[1]);
2743 unixctl_command_reply_error(conn, "Unknown bridge");
2748 ofproto_get_all_flows(br->ofproto, &results);
2750 unixctl_command_reply(conn, ds_cstr(&results));
2751 ds_destroy(&results);
2754 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2755 * connections and reconnect. If BRIDGE is not specified, then all bridges
2756 * drop their controller connections and reconnect. */
2758 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2759 const char *argv[], void *aux OVS_UNUSED)
2763 br = bridge_lookup(argv[1]);
2765 unixctl_command_reply_error(conn, "Unknown bridge");
2768 ofproto_reconnect_controllers(br->ofproto);
2770 HMAP_FOR_EACH (br, node, &all_bridges) {
2771 ofproto_reconnect_controllers(br->ofproto);
2774 unixctl_command_reply(conn, NULL);
2778 bridge_get_controllers(const struct bridge *br,
2779 struct ovsrec_controller ***controllersp)
2781 struct ovsrec_controller **controllers;
2782 size_t n_controllers;
2784 controllers = br->cfg->controller;
2785 n_controllers = br->cfg->n_controller;
2787 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2793 *controllersp = controllers;
2795 return n_controllers;
2799 bridge_queue_if_cfg(struct bridge *br,
2800 const struct ovsrec_interface *cfg,
2801 const struct ovsrec_port *parent)
2803 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2806 if_cfg->parent = parent;
2807 if_cfg->ofport = iface_pick_ofport(cfg);
2808 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2809 hash_string(if_cfg->cfg->name, 0));
2812 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2813 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2814 * 'br' needs to complete its configuration. */
2816 bridge_add_del_ports(struct bridge *br,
2817 const unsigned long int *splinter_vlans)
2819 struct shash_node *port_node;
2820 struct port *port, *next;
2821 struct shash new_ports;
2824 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2826 /* Collect new ports. */
2827 shash_init(&new_ports);
2828 for (i = 0; i < br->cfg->n_ports; i++) {
2829 const char *name = br->cfg->ports[i]->name;
2830 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2831 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2835 if (bridge_get_controllers(br, NULL)
2836 && !shash_find(&new_ports, br->name)) {
2837 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2838 br->name, br->name);
2840 ovsrec_interface_init(&br->synth_local_iface);
2841 ovsrec_port_init(&br->synth_local_port);
2843 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2844 br->synth_local_port.n_interfaces = 1;
2845 br->synth_local_port.name = br->name;
2847 br->synth_local_iface.name = br->name;
2848 br->synth_local_iface.type = "internal";
2850 br->synth_local_ifacep = &br->synth_local_iface;
2852 shash_add(&new_ports, br->name, &br->synth_local_port);
2855 if (splinter_vlans) {
2856 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2859 /* Get rid of deleted ports.
2860 * Get rid of deleted interfaces on ports that still exist. */
2861 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2862 port->cfg = shash_find_data(&new_ports, port->name);
2866 port_del_ifaces(port);
2870 /* Update iface->cfg and iface->type in interfaces that still exist.
2871 * Add new interfaces to creation queue. */
2872 SHASH_FOR_EACH (port_node, &new_ports) {
2873 const struct ovsrec_port *port = port_node->data;
2876 for (i = 0; i < port->n_interfaces; i++) {
2877 const struct ovsrec_interface *cfg = port->interfaces[i];
2878 struct iface *iface = iface_lookup(br, cfg->name);
2879 const char *type = iface_get_type(cfg, br->cfg);
2884 } else if (!strcmp(type, "null")) {
2885 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2886 " may be removed in February 2013. Please email"
2887 " dev@openvswitch.org with concerns.",
2890 bridge_queue_if_cfg(br, cfg, port);
2895 shash_destroy(&new_ports);
2898 /* Initializes 'oc' appropriately as a management service controller for
2901 * The caller must free oc->target when it is no longer needed. */
2903 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2904 struct ofproto_controller *oc)
2906 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2907 oc->max_backoff = 0;
2908 oc->probe_interval = 60;
2909 oc->band = OFPROTO_OUT_OF_BAND;
2911 oc->burst_limit = 0;
2912 oc->enable_async_msgs = true;
2915 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2917 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2918 struct ofproto_controller *oc)
2922 oc->target = c->target;
2923 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2924 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2925 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2926 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2927 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2928 oc->burst_limit = (c->controller_burst_limit
2929 ? *c->controller_burst_limit : 0);
2930 oc->enable_async_msgs = (!c->enable_async_messages
2931 || *c->enable_async_messages);
2932 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2933 if (dscp < 0 || dscp > 63) {
2934 dscp = DSCP_DEFAULT;
2939 /* Configures the IP stack for 'br''s local interface properly according to the
2940 * configuration in 'c'. */
2942 bridge_configure_local_iface_netdev(struct bridge *br,
2943 struct ovsrec_controller *c)
2945 struct netdev *netdev;
2946 struct in_addr mask, gateway;
2948 struct iface *local_iface;
2951 /* If there's no local interface or no IP address, give up. */
2952 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2953 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2957 /* Bring up the local interface. */
2958 netdev = local_iface->netdev;
2959 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2961 /* Configure the IP address and netmask. */
2962 if (!c->local_netmask
2963 || !inet_aton(c->local_netmask, &mask)
2965 mask.s_addr = guess_netmask(ip.s_addr);
2967 if (!netdev_set_in4(netdev, ip, mask)) {
2968 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2969 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2972 /* Configure the default gateway. */
2973 if (c->local_gateway
2974 && inet_aton(c->local_gateway, &gateway)
2975 && gateway.s_addr) {
2976 if (!netdev_add_router(netdev, gateway)) {
2977 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2978 br->name, IP_ARGS(gateway.s_addr));
2983 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2984 * in either string are treated as equal to any number of slashes in the other,
2985 * e.g. "x///y" is equal to "x/y".
2987 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2988 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2989 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2990 * 'b' against a prefix of 'a'.
2993 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2995 const char *b_start = b;
2997 if (b - b_start >= b_stoplen) {
2999 } else if (*a != *b) {
3001 } else if (*a == '/') {
3002 a += strspn(a, "/");
3003 b += strspn(b, "/");
3004 } else if (*a == '\0') {
3014 bridge_configure_remotes(struct bridge *br,
3015 const struct sockaddr_in *managers, size_t n_managers)
3017 bool disable_in_band;
3019 struct ovsrec_controller **controllers;
3020 size_t n_controllers;
3022 enum ofproto_fail_mode fail_mode;
3024 struct ofproto_controller *ocs;
3028 /* Check if we should disable in-band control on this bridge. */
3029 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3032 /* Set OpenFlow queue ID for in-band control. */
3033 ofproto_set_in_band_queue(br->ofproto,
3034 smap_get_int(&br->cfg->other_config,
3035 "in-band-queue", -1));
3037 if (disable_in_band) {
3038 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3040 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3043 n_controllers = bridge_get_controllers(br, &controllers);
3045 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3048 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3049 for (i = 0; i < n_controllers; i++) {
3050 struct ovsrec_controller *c = controllers[i];
3052 if (!strncmp(c->target, "punix:", 6)
3053 || !strncmp(c->target, "unix:", 5)) {
3054 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3057 if (!strncmp(c->target, "unix:", 5)) {
3058 /* Connect to a listening socket */
3059 whitelist = xasprintf("unix:%s/", ovs_rundir());
3060 if (strchr(c->target, '/') &&
3061 !equal_pathnames(c->target, whitelist,
3062 strlen(whitelist))) {
3063 /* Absolute path specified, but not in ovs_rundir */
3064 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3065 "controller \"%s\" due to possibility for "
3066 "remote exploit. Instead, specify socket "
3067 "in whitelisted \"%s\" or connect to "
3068 "\"unix:%s/%s.mgmt\" (which is always "
3069 "available without special configuration).",
3070 br->name, c->target, whitelist,
3071 ovs_rundir(), br->name);
3076 whitelist = xasprintf("punix:%s/%s.controller",
3077 ovs_rundir(), br->name);
3078 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3079 /* Prevent remote ovsdb-server users from accessing
3080 * arbitrary Unix domain sockets and overwriting arbitrary
3082 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3083 "controller \"%s\" due to possibility of "
3084 "overwriting local files. Instead, specify "
3085 "whitelisted \"%s\" or connect to "
3086 "\"unix:%s/%s.mgmt\" (which is always "
3087 "available without special configuration).",
3088 br->name, c->target, whitelist,
3089 ovs_rundir(), br->name);
3098 bridge_configure_local_iface_netdev(br, c);
3099 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3100 if (disable_in_band) {
3101 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3106 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3107 bridge_get_allowed_versions(br));
3108 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3111 /* Set the fail-mode. */
3112 fail_mode = !br->cfg->fail_mode
3113 || !strcmp(br->cfg->fail_mode, "standalone")
3114 ? OFPROTO_FAIL_STANDALONE
3115 : OFPROTO_FAIL_SECURE;
3116 ofproto_set_fail_mode(br->ofproto, fail_mode);
3118 /* Configure OpenFlow controller connection snooping. */
3119 if (!ofproto_has_snoops(br->ofproto)) {
3123 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3124 ovs_rundir(), br->name));
3125 ofproto_set_snoops(br->ofproto, &snoops);
3126 sset_destroy(&snoops);
3131 bridge_configure_tables(struct bridge *br)
3133 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3137 n_tables = ofproto_get_n_tables(br->ofproto);
3139 for (i = 0; i < n_tables; i++) {
3140 struct ofproto_table_settings s;
3143 s.max_flows = UINT_MAX;
3147 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3148 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3151 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3152 s.max_flows = *cfg->flow_limit;
3154 if (cfg->overflow_policy
3155 && !strcmp(cfg->overflow_policy, "evict")) {
3158 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3159 for (k = 0; k < cfg->n_groups; k++) {
3160 const char *string = cfg->groups[k];
3163 msg = mf_parse_subfield__(&s.groups[k], &string);
3165 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3166 "'groups' (%s)", br->name, i, msg);
3168 } else if (*string) {
3169 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3170 "element '%s' contains trailing garbage",
3171 br->name, i, cfg->groups[k]);
3179 ofproto_configure_table(br->ofproto, i, &s);
3183 for (; j < br->cfg->n_flow_tables; j++) {
3184 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3185 "%"PRId64" not supported by this datapath", br->name,
3186 br->cfg->key_flow_tables[j]);
3191 bridge_configure_dp_desc(struct bridge *br)
3193 ofproto_set_dp_desc(br->ofproto,
3194 smap_get(&br->cfg->other_config, "dp-desc"));
3197 /* Port functions. */
3199 static struct port *
3200 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3204 port = xzalloc(sizeof *port);
3206 port->name = xstrdup(cfg->name);
3208 list_init(&port->ifaces);
3210 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3214 /* Deletes interfaces from 'port' that are no longer configured for it. */
3216 port_del_ifaces(struct port *port)
3218 struct iface *iface, *next;
3219 struct sset new_ifaces;
3222 /* Collect list of new interfaces. */
3223 sset_init(&new_ifaces);
3224 for (i = 0; i < port->cfg->n_interfaces; i++) {
3225 const char *name = port->cfg->interfaces[i]->name;
3226 const char *type = port->cfg->interfaces[i]->type;
3227 if (strcmp(type, "null")) {
3228 sset_add(&new_ifaces, name);
3232 /* Get rid of deleted interfaces. */
3233 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3234 if (!sset_contains(&new_ifaces, iface->name)) {
3235 iface_destroy(iface);
3239 sset_destroy(&new_ifaces);
3243 port_destroy(struct port *port)
3246 struct bridge *br = port->bridge;
3247 struct iface *iface, *next;
3250 ofproto_bundle_unregister(br->ofproto, port);
3253 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3254 iface_destroy(iface);
3257 hmap_remove(&br->ports, &port->hmap_node);
3263 static struct port *
3264 port_lookup(const struct bridge *br, const char *name)
3268 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3270 if (!strcmp(port->name, name)) {
3278 enable_lacp(struct port *port, bool *activep)
3280 if (!port->cfg->lacp) {
3281 /* XXX when LACP implementation has been sufficiently tested, enable by
3282 * default and make active on bonded ports. */
3284 } else if (!strcmp(port->cfg->lacp, "off")) {
3286 } else if (!strcmp(port->cfg->lacp, "active")) {
3289 } else if (!strcmp(port->cfg->lacp, "passive")) {
3293 VLOG_WARN("port %s: unknown LACP mode %s",
3294 port->name, port->cfg->lacp);
3299 static struct lacp_settings *
3300 port_configure_lacp(struct port *port, struct lacp_settings *s)
3302 const char *lacp_time, *system_id;
3305 if (!enable_lacp(port, &s->active)) {
3309 s->name = port->name;
3311 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3313 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3314 ETH_ADDR_SCAN_ARGS(s->id))) {
3315 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3316 " address.", port->name, system_id);
3320 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3323 if (eth_addr_is_zero(s->id)) {
3324 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3328 /* Prefer bondable links if unspecified. */
3329 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3331 s->priority = (priority > 0 && priority <= UINT16_MAX
3333 : UINT16_MAX - !list_is_short(&port->ifaces));
3335 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3336 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3338 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3339 "lacp-fallback-ab", false);
3345 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3347 int priority, portid, key;
3349 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3350 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3352 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3354 if (portid <= 0 || portid > UINT16_MAX) {
3355 portid = ofp_to_u16(iface->ofp_port);
3358 if (priority <= 0 || priority > UINT16_MAX) {
3359 priority = UINT16_MAX;
3362 if (key < 0 || key > UINT16_MAX) {
3366 s->name = iface->name;
3368 s->priority = priority;
3373 port_configure_bond(struct port *port, struct bond_settings *s)
3375 const char *detect_s;
3376 struct iface *iface;
3377 int miimon_interval;
3379 s->name = port->name;
3381 if (port->cfg->bond_mode) {
3382 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3383 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3384 port->name, port->cfg->bond_mode,
3385 bond_mode_to_string(s->balance));
3388 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3390 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3391 * active-backup. At some point we should remove this warning. */
3392 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3393 " in previous versions, the default bond_mode was"
3394 " balance-slb", port->name,
3395 bond_mode_to_string(s->balance));
3397 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3398 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3399 "please use another bond type or disable flood_vlans",
3403 miimon_interval = smap_get_int(&port->cfg->other_config,
3404 "bond-miimon-interval", 0);
3405 if (miimon_interval <= 0) {
3406 miimon_interval = 200;
3409 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3410 if (!detect_s || !strcmp(detect_s, "carrier")) {
3411 miimon_interval = 0;
3412 } else if (strcmp(detect_s, "miimon")) {
3413 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3414 "defaulting to carrier", port->name, detect_s);
3415 miimon_interval = 0;
3418 s->up_delay = MAX(0, port->cfg->bond_updelay);
3419 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3420 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3421 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3422 "bond-rebalance-interval", 10000);
3423 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3424 s->rebalance_interval = 1000;
3427 s->fake_iface = port->cfg->bond_fake_iface;
3429 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3430 "lacp-fallback-ab", false);
3432 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3433 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3437 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3438 * instead of obtaining it from the database. */
3440 port_is_synthetic(const struct port *port)
3442 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3445 /* Interface functions. */
3448 iface_is_internal(const struct ovsrec_interface *iface,
3449 const struct ovsrec_bridge *br)
3451 /* The local port and "internal" ports are always "internal". */
3452 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3455 /* Returns the correct network device type for interface 'iface' in bridge
3458 iface_get_type(const struct ovsrec_interface *iface,
3459 const struct ovsrec_bridge *br)
3463 /* The local port always has type "internal". Other ports take
3464 * their type from the database and default to "system" if none is
3466 if (iface_is_internal(iface, br)) {
3469 type = iface->type[0] ? iface->type : "system";
3472 return ofproto_port_open_type(br->datapath_type, type);
3476 iface_destroy(struct iface *iface)
3479 struct port *port = iface->port;
3480 struct bridge *br = port->bridge;
3482 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3483 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3486 if (iface->ofp_port != OFPP_NONE) {
3487 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3490 list_remove(&iface->port_elem);
3491 hmap_remove(&br->iface_by_name, &iface->name_node);
3493 netdev_close(iface->netdev);
3500 static struct iface *
3501 iface_lookup(const struct bridge *br, const char *name)
3503 struct iface *iface;
3505 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3506 &br->iface_by_name) {
3507 if (!strcmp(iface->name, name)) {
3515 static struct iface *
3516 iface_find(const char *name)
3518 const struct bridge *br;
3520 HMAP_FOR_EACH (br, node, &all_bridges) {
3521 struct iface *iface = iface_lookup(br, name);
3530 static struct if_cfg *
3531 if_cfg_lookup(const struct bridge *br, const char *name)
3533 struct if_cfg *if_cfg;
3535 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3537 if (!strcmp(if_cfg->cfg->name, name)) {
3545 static struct iface *
3546 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3548 struct iface *iface;
3550 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3552 if (iface->ofp_port == ofp_port) {
3559 /* Set Ethernet address of 'iface', if one is specified in the configuration
3562 iface_set_mac(struct iface *iface)
3564 uint8_t ea[ETH_ADDR_LEN];
3566 if (!strcmp(iface->type, "internal")
3567 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3568 if (iface->ofp_port == OFPP_LOCAL) {
3569 VLOG_ERR("interface %s: ignoring mac in Interface record "
3570 "(use Bridge record to set local port's mac)",
3572 } else if (eth_addr_is_multicast(ea)) {
3573 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3576 int error = netdev_set_etheraddr(iface->netdev, ea);
3578 VLOG_ERR("interface %s: setting MAC failed (%s)",
3579 iface->name, ovs_strerror(error));
3585 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3587 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3590 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3591 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3592 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3596 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3597 * sets the "ofport" field to -1.
3599 * This is appropriate when 'if_cfg''s interface cannot be created or is
3600 * otherwise invalid. */
3602 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3604 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3605 ovsrec_interface_set_status(if_cfg, NULL);
3606 ovsrec_interface_set_admin_state(if_cfg, NULL);
3607 ovsrec_interface_set_duplex(if_cfg, NULL);
3608 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3609 ovsrec_interface_set_link_state(if_cfg, NULL);
3610 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3611 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3612 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3613 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3614 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3615 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3616 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3617 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3622 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3624 union ovsdb_atom atom;
3626 atom.integer = target;
3627 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3631 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3633 struct ofpbuf queues_buf;
3635 ofpbuf_init(&queues_buf, 0);
3637 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3638 netdev_set_qos(iface->netdev, NULL, NULL);
3640 const struct ovsdb_datum *queues;
3641 struct netdev_queue_dump dump;
3642 unsigned int queue_id;
3643 struct smap details;
3647 /* Configure top-level Qos for 'iface'. */
3648 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3650 /* Deconfigure queues that were deleted. */
3651 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3653 smap_init(&details);
3654 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3655 if (!queue_ids_include(queues, queue_id)) {
3656 netdev_delete_queue(iface->netdev, queue_id);
3659 smap_destroy(&details);
3661 /* Configure queues for 'iface'. */
3663 for (i = 0; i < qos->n_queues; i++) {
3664 const struct ovsrec_queue *queue = qos->value_queues[i];
3665 unsigned int queue_id = qos->key_queues[i];
3667 if (queue_id == 0) {
3671 if (queue->n_dscp == 1) {
3672 struct ofproto_port_queue *port_queue;
3674 port_queue = ofpbuf_put_uninit(&queues_buf,
3675 sizeof *port_queue);
3676 port_queue->queue = queue_id;
3677 port_queue->dscp = queue->dscp[0];
3680 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3683 struct smap details;
3685 smap_init(&details);
3686 netdev_set_queue(iface->netdev, 0, &details);
3687 smap_destroy(&details);
3691 if (iface->ofp_port != OFPP_NONE) {
3692 const struct ofproto_port_queue *port_queues = queues_buf.data;
3693 size_t n_queues = queues_buf.size / sizeof *port_queues;
3695 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3696 port_queues, n_queues);
3699 netdev_set_policing(iface->netdev,
3700 iface->cfg->ingress_policing_rate,
3701 iface->cfg->ingress_policing_burst);
3703 ofpbuf_uninit(&queues_buf);
3707 iface_configure_cfm(struct iface *iface)
3709 const struct ovsrec_interface *cfg = iface->cfg;
3710 const char *opstate_str;
3711 const char *cfm_ccm_vlan;
3712 struct cfm_settings s;
3713 struct smap netdev_args;
3715 if (!cfg->n_cfm_mpid) {
3716 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3720 s.check_tnl_key = false;
3721 smap_init(&netdev_args);
3722 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3723 const char *key = smap_get(&netdev_args, "key");
3724 const char *in_key = smap_get(&netdev_args, "in_key");
3726 s.check_tnl_key = (key && !strcmp(key, "flow"))
3727 || (in_key && !strcmp(in_key, "flow"));
3729 smap_destroy(&netdev_args);
3731 s.mpid = *cfg->cfm_mpid;
3732 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3733 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3734 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3736 if (s.interval <= 0) {
3740 if (!cfm_ccm_vlan) {
3742 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3743 s.ccm_vlan = CFM_RANDOM_VLAN;
3745 s.ccm_vlan = atoi(cfm_ccm_vlan);
3746 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3751 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3753 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3755 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3756 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3758 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3761 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3762 * instead of obtaining it from the database. */
3764 iface_is_synthetic(const struct iface *iface)
3766 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3770 iface_pick_ofport(const struct ovsrec_interface *cfg)
3772 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3774 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3779 /* Port mirroring. */
3781 static struct mirror *
3782 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3786 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3787 if (uuid_equals(uuid, &m->uuid)) {
3795 bridge_configure_mirrors(struct bridge *br)
3797 const struct ovsdb_datum *mc;
3798 unsigned long *flood_vlans;
3799 struct mirror *m, *next;
3802 /* Get rid of deleted mirrors. */
3803 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3804 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3805 union ovsdb_atom atom;
3807 atom.uuid = m->uuid;
3808 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3813 /* Add new mirrors and reconfigure existing ones. */
3814 for (i = 0; i < br->cfg->n_mirrors; i++) {
3815 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3816 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3818 m = mirror_create(br, cfg);
3821 if (!mirror_configure(m)) {
3826 /* Update flooded vlans (for RSPAN). */
3827 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3828 br->cfg->n_flood_vlans);
3829 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3830 bitmap_free(flood_vlans);
3833 static struct mirror *
3834 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3838 m = xzalloc(sizeof *m);
3839 m->uuid = cfg->header_.uuid;
3840 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3842 m->name = xstrdup(cfg->name);
3848 mirror_destroy(struct mirror *m)
3851 struct bridge *br = m->bridge;
3854 ofproto_mirror_unregister(br->ofproto, m);
3857 hmap_remove(&br->mirrors, &m->hmap_node);
3864 mirror_collect_ports(struct mirror *m,
3865 struct ovsrec_port **in_ports, int n_in_ports,
3866 void ***out_portsp, size_t *n_out_portsp)
3868 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3869 size_t n_out_ports = 0;
3872 for (i = 0; i < n_in_ports; i++) {
3873 const char *name = in_ports[i]->name;
3874 struct port *port = port_lookup(m->bridge, name);
3876 out_ports[n_out_ports++] = port;
3878 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3879 "port %s", m->bridge->name, m->name, name);
3882 *out_portsp = out_ports;
3883 *n_out_portsp = n_out_ports;
3887 mirror_configure(struct mirror *m)
3889 const struct ovsrec_mirror *cfg = m->cfg;
3890 struct ofproto_mirror_settings s;
3893 if (strcmp(cfg->name, m->name)) {
3895 m->name = xstrdup(cfg->name);
3899 /* Get output port or VLAN. */
3900 if (cfg->output_port) {
3901 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3902 if (!s.out_bundle) {
3903 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3904 m->bridge->name, m->name);
3907 s.out_vlan = UINT16_MAX;
3909 if (cfg->output_vlan) {
3910 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3911 "output vlan; ignoring output vlan",
3912 m->bridge->name, m->name);
3914 } else if (cfg->output_vlan) {
3915 /* The database should prevent invalid VLAN values. */
3916 s.out_bundle = NULL;
3917 s.out_vlan = *cfg->output_vlan;
3919 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3920 m->bridge->name, m->name);
3924 /* Get port selection. */
3925 if (cfg->select_all) {
3926 size_t n_ports = hmap_count(&m->bridge->ports);
3927 void **ports = xmalloc(n_ports * sizeof *ports);
3932 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3942 /* Get ports, dropping ports that don't exist.
3943 * The IDL ensures that there are no duplicates. */
3944 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3945 &s.srcs, &s.n_srcs);
3946 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3947 &s.dsts, &s.n_dsts);
3950 /* Get VLAN selection. */
3951 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3954 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3957 if (s.srcs != s.dsts) {
3966 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3968 * This is deprecated. It is only for compatibility with broken device drivers
3969 * in old versions of Linux that do not properly support VLANs when VLAN
3970 * devices are not used. When broken device drivers are no longer in
3971 * widespread use, we will delete these interfaces. */
3973 static struct ovsrec_port **recs;
3974 static size_t n_recs, allocated_recs;
3976 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3977 * splinters are reconfigured. */
3979 register_rec(struct ovsrec_port *rec)
3981 if (n_recs >= allocated_recs) {
3982 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3984 recs[n_recs++] = rec;
3987 /* Frees all of the ports registered with register_reg(). */
3989 free_registered_recs(void)
3993 for (i = 0; i < n_recs; i++) {
3994 struct ovsrec_port *port = recs[i];
3997 for (j = 0; j < port->n_interfaces; j++) {
3998 struct ovsrec_interface *iface = port->interfaces[j];
4003 smap_destroy(&port->other_config);
4004 free(port->interfaces);
4012 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4015 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4017 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4021 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4024 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4025 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4026 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4029 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4030 * use, returns NULL.
4032 * Updates 'vlan_splinters_enabled_anywhere'. */
4033 static unsigned long int *
4034 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4036 unsigned long int *splinter_vlans;
4037 struct sset splinter_ifaces;
4038 const char *real_dev_name;
4039 struct shash *real_devs;
4040 struct shash_node *node;
4044 /* Free space allocated for synthesized ports and interfaces, since we're
4045 * in the process of reconstructing all of them. */
4046 free_registered_recs();
4048 splinter_vlans = bitmap_allocate(4096);
4049 sset_init(&splinter_ifaces);
4050 vlan_splinters_enabled_anywhere = false;
4051 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4052 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4055 for (j = 0; j < br_cfg->n_ports; j++) {
4056 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4059 for (k = 0; k < port_cfg->n_interfaces; k++) {
4060 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4062 if (vlan_splinters_is_enabled(iface_cfg)) {
4063 vlan_splinters_enabled_anywhere = true;
4064 sset_add(&splinter_ifaces, iface_cfg->name);
4065 vlan_bitmap_from_array__(port_cfg->trunks,
4071 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4072 bitmap_set1(splinter_vlans, *port_cfg->tag);
4077 if (!vlan_splinters_enabled_anywhere) {
4078 free(splinter_vlans);
4079 sset_destroy(&splinter_ifaces);
4083 HMAP_FOR_EACH (br, node, &all_bridges) {
4085 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4089 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4090 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4091 * device to be created for it. */
4092 bitmap_set0(splinter_vlans, 0);
4093 bitmap_set0(splinter_vlans, 4095);
4095 /* Delete all VLAN devices that we don't need. */
4097 real_devs = vlandev_get_real_devs();
4098 SHASH_FOR_EACH (node, real_devs) {
4099 const struct vlan_real_dev *real_dev = node->data;
4100 const struct vlan_dev *vlan_dev;
4101 bool real_dev_has_splinters;
4103 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4105 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4106 if (!real_dev_has_splinters
4107 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4108 struct netdev *netdev;
4110 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4111 if (!netdev_get_in4(netdev, NULL, NULL) ||
4112 !netdev_get_in6(netdev, NULL)) {
4113 /* It has an IP address configured, so we don't own
4114 * it. Don't delete it. */
4116 vlandev_del(vlan_dev->name);
4118 netdev_close(netdev);
4125 /* Add all VLAN devices that we need. */
4126 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4129 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4130 if (!vlandev_get_name(real_dev_name, vid)) {
4131 vlandev_add(real_dev_name, vid);
4138 sset_destroy(&splinter_ifaces);
4140 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4141 free(splinter_vlans);
4144 return splinter_vlans;
4147 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4150 configure_splinter_port(struct port *port)
4152 struct ofproto *ofproto = port->bridge->ofproto;
4153 ofp_port_t realdev_ofp_port;
4154 const char *realdev_name;
4155 struct iface *vlandev, *realdev;
4157 ofproto_bundle_unregister(port->bridge->ofproto, port);
4159 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4162 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4163 realdev = iface_lookup(port->bridge, realdev_name);
4164 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4166 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4170 static struct ovsrec_port *
4171 synthesize_splinter_port(const char *real_dev_name,
4172 const char *vlan_dev_name, int vid)
4174 struct ovsrec_interface *iface;
4175 struct ovsrec_port *port;
4177 iface = xmalloc(sizeof *iface);
4178 ovsrec_interface_init(iface);
4179 iface->name = xstrdup(vlan_dev_name);
4180 iface->type = "system";
4182 port = xmalloc(sizeof *port);
4183 ovsrec_port_init(port);
4184 port->interfaces = xmemdup(&iface, sizeof iface);
4185 port->n_interfaces = 1;
4186 port->name = xstrdup(vlan_dev_name);
4187 port->vlan_mode = "splinter";
4188 port->tag = xmalloc(sizeof *port->tag);
4191 smap_add(&port->other_config, "realdev", real_dev_name);
4197 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4198 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4199 * 1-bit in the 'splinter_vlans' bitmap. */
4201 add_vlan_splinter_ports(struct bridge *br,
4202 const unsigned long int *splinter_vlans,
4203 struct shash *ports)
4207 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4208 * we're modifying 'ports'. */
4209 for (i = 0; i < br->cfg->n_ports; i++) {
4210 const char *name = br->cfg->ports[i]->name;
4211 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4214 for (j = 0; j < port_cfg->n_interfaces; j++) {
4215 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4217 if (vlan_splinters_is_enabled(iface_cfg)) {
4218 const char *real_dev_name;
4221 real_dev_name = iface_cfg->name;
4222 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4223 const char *vlan_dev_name;
4225 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4227 && !shash_find(ports, vlan_dev_name)) {
4228 shash_add(ports, vlan_dev_name,
4229 synthesize_splinter_port(
4230 real_dev_name, vlan_dev_name, vid));
4239 mirror_refresh_stats(struct mirror *m)
4241 struct ofproto *ofproto = m->bridge->ofproto;
4242 uint64_t tx_packets, tx_bytes;
4245 size_t stat_cnt = 0;
4247 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4248 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4252 if (tx_packets != UINT64_MAX) {
4253 keys[stat_cnt] = "tx_packets";
4254 values[stat_cnt] = tx_packets;
4257 if (tx_bytes != UINT64_MAX) {
4258 keys[stat_cnt] = "tx_bytes";
4259 values[stat_cnt] = tx_bytes;
4263 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);