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.
28 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "meta-flow.h"
38 #include "ofp-print.h"
41 #include "ofproto/ofproto.h"
42 #include "poll-loop.h"
46 #include "socket-util.h"
48 #include "stream-ssl.h"
50 #include "system-stats.h"
55 #include "lib/vswitch-idl.h"
56 #include "xenserver.h"
58 #include "sflow_api.h"
59 #include "vlan-bitmap.h"
61 VLOG_DEFINE_THIS_MODULE(bridge);
63 COVERAGE_DEFINE(bridge_reconfigure);
65 /* Configuration of an uninstantiated iface. */
67 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
68 const struct ovsrec_interface *cfg; /* Interface record. */
69 const struct ovsrec_port *parent; /* Parent port record. */
70 ofp_port_t ofport; /* Requested OpenFlow port number. */
73 /* OpenFlow port slated for removal from ofproto. */
75 struct list list_node; /* Node in bridge's ofpp_garbage. */
76 ofp_port_t ofp_port; /* Port to be deleted. */
80 /* These members are always valid. */
81 struct list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
86 /* These members are valid only after bridge_reconfigure() causes them to
88 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
89 ofp_port_t ofp_port; /* OpenFlow port number, */
90 /* OFPP_NONE if unknown. */
91 struct netdev *netdev; /* Network device. */
92 const char *type; /* Usually same as cfg->type. */
93 const struct ovsrec_interface *cfg;
97 struct uuid uuid; /* UUID of this "mirror" record in database. */
98 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
99 struct bridge *bridge;
101 const struct ovsrec_mirror *cfg;
105 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
106 struct bridge *bridge;
109 const struct ovsrec_port *cfg;
111 /* An ordinary bridge port has 1 interface.
112 * A bridge port for bonding has at least 2 interfaces. */
113 struct list ifaces; /* List of "struct iface"s. */
117 struct hmap_node node; /* In 'all_bridges'. */
118 char *name; /* User-specified arbitrary name. */
119 char *type; /* Datapath type. */
120 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
121 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
122 const struct ovsrec_bridge *cfg;
124 /* OpenFlow switch processing. */
125 struct ofproto *ofproto; /* OpenFlow switch. */
128 struct hmap ports; /* "struct port"s indexed by name. */
129 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
130 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
132 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
133 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
134 Indexed on 'cfg->name'. */
136 /* Port mirroring. */
137 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
145 /* All bridges, indexed by name. */
146 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
148 /* OVSDB IDL used to obtain configuration. */
149 static struct ovsdb_idl *idl;
151 /* We want to complete daemonization, fully detaching from our parent process,
152 * only after we have completed our initial configuration, committed our state
153 * to the database, and received confirmation back from the database server
154 * that it applied the commit. This allows our parent process to know that,
155 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
156 * to have already been filled in. (It is only "very likely" rather than
157 * certain because there is always a slim possibility that the transaction will
158 * fail or that some other client has added new bridges, ports, etc. while
159 * ovs-vswitchd was configuring using an old configuration.)
161 * We only need to do this once for our initial configuration at startup, so
162 * 'initial_config_done' tracks whether we've already done it. While we are
163 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
164 * itself and is otherwise NULL. */
165 static bool initial_config_done;
166 static struct ovsdb_idl_txn *daemonize_txn;
168 /* Most recently processed IDL sequence number. */
169 static unsigned int idl_seqno;
171 /* Each time this timer expires, the bridge fetches interface and mirror
172 * statistics and pushes them into the database. */
173 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
174 static long long int iface_stats_timer = LLONG_MIN;
176 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
177 * expensive. This can cause bridge_reconfigure() to take a long time during
178 * which no other work can be done. To deal with this problem, we limit port
179 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
180 * call to bridge_reconfigure(). If there is more work to do after the limit
181 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
182 * This allows the rest of the code to catch up on important things like
183 * forwarding packets. */
184 #define OFP_PORT_ACTION_WINDOW 10
185 static bool reconfiguring = false;
187 static void add_del_bridges(const struct ovsrec_open_vswitch *);
188 static void bridge_update_ofprotos(void);
189 static void bridge_create(const struct ovsrec_bridge *);
190 static void bridge_destroy(struct bridge *);
191 static struct bridge *bridge_lookup(const char *name);
192 static unixctl_cb_func bridge_unixctl_dump_flows;
193 static unixctl_cb_func bridge_unixctl_reconnect;
194 static size_t bridge_get_controllers(const struct bridge *br,
195 struct ovsrec_controller ***controllersp);
196 static void bridge_add_del_ports(struct bridge *,
197 const unsigned long int *splinter_vlans);
198 static void bridge_refresh_ofp_port(struct bridge *);
199 static void bridge_configure_flow_miss_model(const char *opt);
200 static void bridge_configure_datapath_id(struct bridge *);
201 static void bridge_configure_netflow(struct bridge *);
202 static void bridge_configure_forward_bpdu(struct bridge *);
203 static void bridge_configure_mac_table(struct bridge *);
204 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
205 static void bridge_configure_ipfix(struct bridge *);
206 static void bridge_configure_stp(struct bridge *);
207 static void bridge_configure_tables(struct bridge *);
208 static void bridge_configure_dp_desc(struct bridge *);
209 static void bridge_configure_remotes(struct bridge *,
210 const struct sockaddr_in *managers,
212 static void bridge_pick_local_hw_addr(struct bridge *,
213 uint8_t ea[ETH_ADDR_LEN],
214 struct iface **hw_addr_iface);
215 static uint64_t bridge_pick_datapath_id(struct bridge *,
216 const uint8_t bridge_ea[ETH_ADDR_LEN],
217 struct iface *hw_addr_iface);
218 static void bridge_queue_if_cfg(struct bridge *,
219 const struct ovsrec_interface *,
220 const struct ovsrec_port *);
221 static uint64_t dpid_from_hash(const void *, size_t nbytes);
222 static bool bridge_has_bond_fake_iface(const struct bridge *,
224 static bool port_is_bond_fake_iface(const struct port *);
226 static unixctl_cb_func qos_unixctl_show;
228 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
229 static void port_del_ifaces(struct port *);
230 static void port_destroy(struct port *);
231 static struct port *port_lookup(const struct bridge *, const char *name);
232 static void port_configure(struct port *);
233 static struct lacp_settings *port_configure_lacp(struct port *,
234 struct lacp_settings *);
235 static void port_configure_bond(struct port *, struct bond_settings *);
236 static bool port_is_synthetic(const struct port *);
238 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
239 static void run_system_stats(void);
241 static void bridge_configure_mirrors(struct bridge *);
242 static struct mirror *mirror_create(struct bridge *,
243 const struct ovsrec_mirror *);
244 static void mirror_destroy(struct mirror *);
245 static bool mirror_configure(struct mirror *);
246 static void mirror_refresh_stats(struct mirror *);
248 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
249 static bool iface_create(struct bridge *, struct if_cfg *,
250 ofp_port_t ofp_port);
251 static bool iface_is_internal(const struct ovsrec_interface *iface,
252 const struct ovsrec_bridge *br);
253 static const char *iface_get_type(const struct ovsrec_interface *,
254 const struct ovsrec_bridge *);
255 static void iface_destroy(struct iface *);
256 static struct iface *iface_lookup(const struct bridge *, const char *name);
257 static struct iface *iface_find(const char *name);
258 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
259 static struct iface *iface_from_ofp_port(const struct bridge *,
260 ofp_port_t ofp_port);
261 static void iface_set_mac(struct iface *);
262 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
263 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
264 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
265 static void iface_configure_cfm(struct iface *);
266 static void iface_refresh_cfm_stats(struct iface *);
267 static void iface_refresh_stats(struct iface *);
268 static void iface_refresh_status(struct iface *);
269 static bool iface_is_synthetic(const struct iface *);
270 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
272 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
274 * This is deprecated. It is only for compatibility with broken device drivers
275 * in old versions of Linux that do not properly support VLANs when VLAN
276 * devices are not used. When broken device drivers are no longer in
277 * widespread use, we will delete these interfaces. */
279 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
280 static bool vlan_splinters_enabled_anywhere;
282 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
283 static unsigned long int *collect_splinter_vlans(
284 const struct ovsrec_open_vswitch *);
285 static void configure_splinter_port(struct port *);
286 static void add_vlan_splinter_ports(struct bridge *,
287 const unsigned long int *splinter_vlans,
288 struct shash *ports);
291 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
293 struct shash iface_hints;
294 static bool initialized = false;
301 shash_init(&iface_hints);
304 for (i = 0; i < cfg->n_bridges; i++) {
305 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
308 for (j = 0; j < br_cfg->n_ports; j++) {
309 struct ovsrec_port *port_cfg = br_cfg->ports[j];
312 for (k = 0; k < port_cfg->n_interfaces; k++) {
313 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
314 struct iface_hint *iface_hint;
316 iface_hint = xmalloc(sizeof *iface_hint);
317 iface_hint->br_name = br_cfg->name;
318 iface_hint->br_type = br_cfg->datapath_type;
319 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
321 shash_add(&iface_hints, if_cfg->name, iface_hint);
327 ofproto_init(&iface_hints);
329 shash_destroy_free_data(&iface_hints);
333 /* Public functions. */
335 /* Initializes the bridge module, configuring it to obtain its configuration
336 * from an OVSDB server accessed over 'remote', which should be a string in a
337 * form acceptable to ovsdb_idl_create(). */
339 bridge_init(const char *remote)
341 /* Create connection to database. */
342 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
343 idl_seqno = ovsdb_idl_get_seqno(idl);
344 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
345 ovsdb_idl_verify_write_only(idl);
347 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
355 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
357 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
359 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
361 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
381 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
386 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
388 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
393 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
395 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
406 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
408 /* Register unixctl commands. */
409 unixctl_command_register("qos/show", "interface", 1, 1,
410 qos_unixctl_show, NULL);
411 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
412 bridge_unixctl_dump_flows, NULL);
413 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
414 bridge_unixctl_reconnect, NULL);
424 struct bridge *br, *next_br;
426 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
429 ovsdb_idl_destroy(idl);
432 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
433 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
434 * responsible for freeing '*managersp' (with free()).
436 * You may be asking yourself "why does ovs-vswitchd care?", because
437 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
438 * should not be and in fact is not directly involved in that. But
439 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
440 * it has to tell in-band control where the managers are to enable that.
441 * (Thus, only managers connected in-band are collected.)
444 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
445 struct sockaddr_in **managersp, size_t *n_managersp)
447 struct sockaddr_in *managers = NULL;
448 size_t n_managers = 0;
452 /* Collect all of the potential targets from the "targets" columns of the
453 * rows pointed to by "manager_options", excluding any that are
456 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
457 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
459 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
460 sset_find_and_delete(&targets, m->target);
462 sset_add(&targets, m->target);
466 /* Now extract the targets' IP addresses. */
467 if (!sset_is_empty(&targets)) {
470 managers = xmalloc(sset_count(&targets) * sizeof *managers);
471 SSET_FOR_EACH (target, &targets) {
472 struct sockaddr_in *sin = &managers[n_managers];
474 if (stream_parse_target_with_default_ports(target,
482 sset_destroy(&targets);
484 *managersp = managers;
485 *n_managersp = n_managers;
489 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
491 unsigned long int *splinter_vlans;
494 COVERAGE_INC(bridge_reconfigure);
496 ovs_assert(!reconfiguring);
497 reconfiguring = true;
499 ofproto_set_flow_eviction_threshold(
500 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
501 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
503 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
504 "force-miss-model"));
506 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
507 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
508 * configuration otherwise.
510 * This is mostly an update to bridge data structures. Nothing is pushed
511 * down to ofproto or lower layers. */
512 add_del_bridges(ovs_cfg);
513 splinter_vlans = collect_splinter_vlans(ovs_cfg);
514 HMAP_FOR_EACH (br, node, &all_bridges) {
515 bridge_add_del_ports(br, splinter_vlans);
517 free(splinter_vlans);
519 /* Delete datapaths that are no longer configured, and create ones which
520 * don't exist but should. */
521 bridge_update_ofprotos();
523 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
524 HMAP_FOR_EACH (br, node, &all_bridges) {
525 bridge_refresh_ofp_port(br);
528 /* Clear database records for "if_cfg"s which haven't been instantiated. */
529 HMAP_FOR_EACH (br, node, &all_bridges) {
530 struct if_cfg *if_cfg;
532 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
533 iface_clear_db_record(if_cfg->cfg);
537 reconfigure_system_stats(ovs_cfg);
541 bridge_reconfigure_ofp(void)
543 long long int deadline;
547 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
549 /* The kernel will reject any attempt to add a given port to a datapath if
550 * that port already belongs to a different datapath, so we must do all
551 * port deletions before any port additions. */
552 HMAP_FOR_EACH (br, node, &all_bridges) {
553 struct ofpp_garbage *garbage, *next;
555 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
556 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
557 * internal datastructures may not be consistent. Eventually, when
558 * port additions and deletions are cheaper, these calls should be
561 ofproto_port_del(br->ofproto, garbage->ofp_port);
562 list_remove(&garbage->list_node);
566 if (time_msec() >= deadline) {
573 HMAP_FOR_EACH (br, node, &all_bridges) {
574 struct if_cfg *if_cfg, *next;
576 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
577 iface_create(br, if_cfg, OFPP_NONE);
579 if (time_msec() >= deadline) {
589 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
591 struct sockaddr_in *managers;
592 int sflow_bridge_number;
597 ovs_assert(reconfiguring);
598 done = bridge_reconfigure_ofp();
600 /* Complete the configuration. */
601 sflow_bridge_number = 0;
602 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
603 HMAP_FOR_EACH (br, node, &all_bridges) {
606 /* We need the datapath ID early to allow LACP ports to use it as the
607 * default system ID. */
608 bridge_configure_datapath_id(br);
610 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
613 port_configure(port);
615 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
616 iface_configure_cfm(iface);
617 iface_configure_qos(iface, port->cfg->qos);
618 iface_set_mac(iface);
619 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
623 bridge_configure_mirrors(br);
624 bridge_configure_forward_bpdu(br);
625 bridge_configure_mac_table(br);
626 bridge_configure_remotes(br, managers, n_managers);
627 bridge_configure_netflow(br);
628 bridge_configure_sflow(br, &sflow_bridge_number);
629 bridge_configure_ipfix(br);
630 bridge_configure_stp(br);
631 bridge_configure_tables(br);
632 bridge_configure_dp_desc(br);
634 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
635 /* XXX: Remove this warning message eventually. */
636 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
637 " moved to the Open_vSwitch table. Ignoring its"
638 " setting in the bridge table.");
646 /* Delete ofprotos which aren't configured or have the wrong type. Create
647 * ofprotos which don't exist but need to. */
649 bridge_update_ofprotos(void)
651 struct bridge *br, *next;
656 /* Delete ofprotos with no bridge or with the wrong type. */
659 ofproto_enumerate_types(&types);
660 SSET_FOR_EACH (type, &types) {
663 ofproto_enumerate_names(type, &names);
664 SSET_FOR_EACH (name, &names) {
665 br = bridge_lookup(name);
666 if (!br || strcmp(type, br->type)) {
667 ofproto_delete(name, type);
671 sset_destroy(&names);
672 sset_destroy(&types);
674 /* Add ofprotos for bridges which don't have one yet. */
675 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
683 /* Remove ports from any datapath with the same name as 'br'. If we
684 * don't do this, creating 'br''s ofproto will fail because a port with
685 * the same name as its local port already exists. */
686 HMAP_FOR_EACH (br2, node, &all_bridges) {
687 struct ofproto_port ofproto_port;
693 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
695 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
697 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
700 ofproto_port_destroy(&ofproto_port);
704 error = ofproto_create(br->name, br->type, &br->ofproto);
706 VLOG_ERR("failed to create bridge %s: %s", br->name,
714 port_configure(struct port *port)
716 const struct ovsrec_port *cfg = port->cfg;
717 struct bond_settings bond_settings;
718 struct lacp_settings lacp_settings;
719 struct ofproto_bundle_settings s;
722 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
723 configure_splinter_port(port);
732 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
733 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
734 s.slaves[s.n_slaves++] = iface->ofp_port;
739 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
743 /* Get VLAN trunks. */
746 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
750 if (cfg->vlan_mode) {
751 if (!strcmp(cfg->vlan_mode, "access")) {
752 s.vlan_mode = PORT_VLAN_ACCESS;
753 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
754 s.vlan_mode = PORT_VLAN_TRUNK;
755 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
756 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
757 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
758 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
760 /* This "can't happen" because ovsdb-server should prevent it. */
761 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
762 s.vlan_mode = PORT_VLAN_TRUNK;
766 s.vlan_mode = PORT_VLAN_ACCESS;
768 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
772 s.vlan_mode = PORT_VLAN_TRUNK;
775 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
778 /* Get LACP settings. */
779 s.lacp = port_configure_lacp(port, &lacp_settings);
783 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
784 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
785 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
788 s.lacp_slaves = NULL;
791 /* Get bond settings. */
792 if (s.n_slaves > 1) {
793 s.bond = &bond_settings;
794 port_configure_bond(port, &bond_settings);
797 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
798 netdev_set_miimon_interval(iface->netdev, 0);
803 ofproto_bundle_register(port->bridge->ofproto, port, &s);
812 bridge_configure_flow_miss_model(const char *opt)
814 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
817 if (strcmp(opt, "with-facets")) {
818 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
819 VLOG_INFO("Handling all flow misses by creating facets.\n");
821 if (strcmp(opt, "without-facets")) {
822 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
823 VLOG_INFO("Handling all flow misses without creating facets.\n");
827 ofproto_set_flow_miss_model(model);
830 /* Pick local port hardware address and datapath ID for 'br'. */
832 bridge_configure_datapath_id(struct bridge *br)
834 uint8_t ea[ETH_ADDR_LEN];
836 struct iface *local_iface;
837 struct iface *hw_addr_iface;
840 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
841 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
843 int error = netdev_set_etheraddr(local_iface->netdev, ea);
845 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
846 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
847 "Ethernet address: %s",
848 br->name, strerror(error));
851 memcpy(br->ea, ea, ETH_ADDR_LEN);
853 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
854 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
855 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
856 ofproto_set_datapath_id(br->ofproto, dpid);
859 dpid_string = xasprintf("%016"PRIx64, dpid);
860 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
864 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
867 bridge_get_allowed_versions(struct bridge *br)
869 if (!br->cfg->n_protocols)
872 return ofputil_versions_from_strings(br->cfg->protocols,
873 br->cfg->n_protocols);
876 /* Set NetFlow configuration on 'br'. */
878 bridge_configure_netflow(struct bridge *br)
880 struct ovsrec_netflow *cfg = br->cfg->netflow;
881 struct netflow_options opts;
884 ofproto_set_netflow(br->ofproto, NULL);
888 memset(&opts, 0, sizeof opts);
890 /* Get default NetFlow configuration from datapath.
891 * Apply overrides from 'cfg'. */
892 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
893 if (cfg->engine_type) {
894 opts.engine_type = *cfg->engine_type;
896 if (cfg->engine_id) {
897 opts.engine_id = *cfg->engine_id;
900 /* Configure active timeout interval. */
901 opts.active_timeout = cfg->active_timeout;
902 if (!opts.active_timeout) {
903 opts.active_timeout = -1;
904 } else if (opts.active_timeout < 0) {
905 VLOG_WARN("bridge %s: active timeout interval set to negative "
906 "value, using default instead (%d seconds)", br->name,
907 NF_ACTIVE_TIMEOUT_DEFAULT);
908 opts.active_timeout = -1;
911 /* Add engine ID to interface number to disambiguate bridgs? */
912 opts.add_id_to_iface = cfg->add_id_to_interface;
913 if (opts.add_id_to_iface) {
914 if (opts.engine_id > 0x7f) {
915 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
916 "another vswitch, choose an engine id less than 128",
919 if (hmap_count(&br->ports) > 508) {
920 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
921 "another port when more than 508 ports are used",
927 sset_init(&opts.collectors);
928 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
931 if (ofproto_set_netflow(br->ofproto, &opts)) {
932 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
934 sset_destroy(&opts.collectors);
937 /* Set sFlow configuration on 'br'. */
939 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
941 const struct ovsrec_sflow *cfg = br->cfg->sflow;
942 struct ovsrec_controller **controllers;
943 struct ofproto_sflow_options oso;
944 size_t n_controllers;
948 ofproto_set_sflow(br->ofproto, NULL);
952 memset(&oso, 0, sizeof oso);
954 sset_init(&oso.targets);
955 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
957 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
959 oso.sampling_rate = *cfg->sampling;
962 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
964 oso.polling_interval = *cfg->polling;
967 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
969 oso.header_len = *cfg->header;
972 oso.sub_id = (*sflow_bridge_number)++;
973 oso.agent_device = cfg->agent;
975 oso.control_ip = NULL;
976 n_controllers = bridge_get_controllers(br, &controllers);
977 for (i = 0; i < n_controllers; i++) {
978 if (controllers[i]->local_ip) {
979 oso.control_ip = controllers[i]->local_ip;
983 ofproto_set_sflow(br->ofproto, &oso);
985 sset_destroy(&oso.targets);
988 /* Set IPFIX configuration on 'br'. */
990 bridge_configure_ipfix(struct bridge *br)
992 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
993 const struct ovsrec_flow_sample_collector_set *fe_cfg;
994 struct ofproto_ipfix_bridge_exporter_options be_opts;
995 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
996 size_t n_fe_opts = 0;
998 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
999 if (fe_cfg->bridge == br->cfg) {
1004 if (!be_cfg && n_fe_opts == 0) {
1005 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1010 memset(&be_opts, 0, sizeof be_opts);
1012 sset_init(&be_opts.targets);
1013 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1015 if (be_cfg->sampling) {
1016 be_opts.sampling_rate = *be_cfg->sampling;
1018 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1020 if (be_cfg->obs_domain_id) {
1021 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1023 if (be_cfg->obs_point_id) {
1024 be_opts.obs_point_id = *be_cfg->obs_point_id;
1028 if (n_fe_opts > 0) {
1029 struct ofproto_ipfix_flow_exporter_options *opts;
1030 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1032 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1033 if (fe_cfg->bridge == br->cfg) {
1034 opts->collector_set_id = fe_cfg->id;
1035 sset_init(&opts->targets);
1036 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1037 fe_cfg->ipfix->n_targets);
1043 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1047 sset_destroy(&be_opts.targets);
1050 if (n_fe_opts > 0) {
1051 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1053 for (i = 0; i < n_fe_opts; i++) {
1054 sset_destroy(&opts->targets);
1062 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1063 struct ofproto_port_stp_settings *port_s,
1064 int *port_num_counter, unsigned long *port_num_bitmap)
1066 const char *config_str;
1067 struct iface *iface;
1069 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1070 port_s->enable = false;
1073 port_s->enable = true;
1076 /* STP over bonds is not supported. */
1077 if (!list_is_singleton(&port->ifaces)) {
1078 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1080 port_s->enable = false;
1084 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1086 /* Internal ports shouldn't participate in spanning tree, so
1088 if (!strcmp(iface->type, "internal")) {
1089 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1090 port_s->enable = false;
1094 /* STP on mirror output ports is not supported. */
1095 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1096 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1097 port_s->enable = false;
1101 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1103 unsigned long int port_num = strtoul(config_str, NULL, 0);
1104 int port_idx = port_num - 1;
1106 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1107 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1108 port_s->enable = false;
1112 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1113 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1114 port->name, port_num);
1115 port_s->enable = false;
1118 bitmap_set1(port_num_bitmap, port_idx);
1119 port_s->port_num = port_idx;
1121 if (*port_num_counter >= STP_MAX_PORTS) {
1122 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1123 port_s->enable = false;
1127 port_s->port_num = (*port_num_counter)++;
1130 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1132 port_s->path_cost = strtoul(config_str, NULL, 10);
1134 enum netdev_features current;
1137 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1138 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1139 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1142 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1144 port_s->priority = strtoul(config_str, NULL, 0);
1146 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1150 /* Set spanning tree configuration on 'br'. */
1152 bridge_configure_stp(struct bridge *br)
1154 if (!br->cfg->stp_enable) {
1155 ofproto_set_stp(br->ofproto, NULL);
1157 struct ofproto_stp_settings br_s;
1158 const char *config_str;
1160 int port_num_counter;
1161 unsigned long *port_num_bitmap;
1163 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1165 uint8_t ea[ETH_ADDR_LEN];
1167 if (eth_addr_from_string(config_str, ea)) {
1168 br_s.system_id = eth_addr_to_uint64(ea);
1170 br_s.system_id = eth_addr_to_uint64(br->ea);
1171 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1172 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1175 br_s.system_id = eth_addr_to_uint64(br->ea);
1178 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1180 br_s.priority = strtoul(config_str, NULL, 0);
1182 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1185 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1187 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1189 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1192 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1194 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1196 br_s.max_age = STP_DEFAULT_MAX_AGE;
1199 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1201 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1203 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1206 /* Configure STP on the bridge. */
1207 if (ofproto_set_stp(br->ofproto, &br_s)) {
1208 VLOG_ERR("bridge %s: could not enable STP", br->name);
1212 /* Users must either set the port number with the "stp-port-num"
1213 * configuration on all ports or none. If manual configuration
1214 * is not done, then we allocate them sequentially. */
1215 port_num_counter = 0;
1216 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1217 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1218 struct ofproto_port_stp_settings port_s;
1219 struct iface *iface;
1221 port_configure_stp(br->ofproto, port, &port_s,
1222 &port_num_counter, port_num_bitmap);
1224 /* As bonds are not supported, just apply configuration to
1225 * all interfaces. */
1226 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1227 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1229 VLOG_ERR("port %s: could not enable STP", port->name);
1235 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1236 && port_num_counter) {
1237 VLOG_ERR("bridge %s: must manually configure all STP port "
1238 "IDs or none, disabling", br->name);
1239 ofproto_set_stp(br->ofproto, NULL);
1241 bitmap_free(port_num_bitmap);
1246 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1248 const struct port *port = port_lookup(br, name);
1249 return port && port_is_bond_fake_iface(port);
1253 port_is_bond_fake_iface(const struct port *port)
1255 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1259 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1261 struct bridge *br, *next;
1262 struct shash new_br;
1265 /* Collect new bridges' names and types. */
1266 shash_init(&new_br);
1267 for (i = 0; i < cfg->n_bridges; i++) {
1268 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1269 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1271 if (strchr(br_cfg->name, '/')) {
1272 /* Prevent remote ovsdb-server users from accessing arbitrary
1273 * directories, e.g. consider a bridge named "../../../etc/". */
1274 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1276 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1277 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1281 /* Get rid of deleted bridges or those whose types have changed.
1282 * Update 'cfg' of bridges that still exist. */
1283 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1284 br->cfg = shash_find_data(&new_br, br->name);
1285 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1286 br->cfg->datapath_type))) {
1291 /* Add new bridges. */
1292 for (i = 0; i < cfg->n_bridges; i++) {
1293 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1294 struct bridge *br = bridge_lookup(br_cfg->name);
1296 bridge_create(br_cfg);
1300 shash_destroy(&new_br);
1304 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1306 struct bridge *br = iface->port->bridge;
1308 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1309 iface->ofp_port = ofp_port;
1310 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1311 hash_ofp_port(ofp_port));
1312 iface_set_ofport(iface->cfg, ofp_port);
1315 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1316 * Returns 0 if successful, otherwise a positive errno value. */
1318 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1319 struct netdev *netdev)
1323 error = netdev_set_config(netdev, &iface_cfg->options);
1325 VLOG_WARN("could not configure network device %s (%s)",
1326 iface_cfg->name, strerror(error));
1331 /* This function determines whether 'ofproto_port', which is attached to
1332 * br->ofproto's datapath, is one that we want in 'br'.
1334 * If it is, it returns true, after creating an iface (if necessary),
1335 * configuring the iface's netdev according to the iface's options, and setting
1336 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1338 * If, on the other hand, 'port' should be removed, it returns false. The
1339 * caller should later detach the port from br->ofproto. */
1341 bridge_refresh_one_ofp_port(struct bridge *br,
1342 const struct ofproto_port *ofproto_port)
1344 const char *name = ofproto_port->name;
1345 const char *type = ofproto_port->type;
1346 ofp_port_t ofp_port = ofproto_port->ofp_port;
1348 struct iface *iface = iface_lookup(br, name);
1350 /* Check that the name-to-number mapping is one-to-one. */
1351 if (iface->ofp_port != OFPP_NONE) {
1352 VLOG_WARN("bridge %s: interface %s reported twice",
1355 } else if (iface_from_ofp_port(br, ofp_port)) {
1356 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1357 br->name, ofp_port);
1361 /* There's a configured interface named 'name'. */
1362 if (strcmp(type, iface->type)
1363 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1364 /* It's the wrong type, or it's the right type but can't be
1365 * configured as the user requested, so we must destroy it. */
1368 /* It's the right type and configured correctly. Keep it. */
1369 iface_set_ofp_port(iface, ofp_port);
1372 } else if (bridge_has_bond_fake_iface(br, name)
1373 && !strcmp(type, "internal")) {
1374 /* It's a bond fake iface. Keep it. */
1377 /* There's no configured interface named 'name', but there might be an
1378 * interface of that name queued to be created.
1380 * If there is, and it has the correct type, then try to configure it
1381 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1382 * delete it and later try to re-add it. */
1383 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1385 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1386 && iface_create(br, if_cfg, ofp_port));
1390 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1391 * consistent with the ofp_ports in "br->ofproto". */
1393 bridge_refresh_ofp_port(struct bridge *br)
1395 struct ofproto_port_dump dump;
1396 struct ofproto_port ofproto_port;
1397 struct port *port, *port_next;
1399 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1400 hmap_clear(&br->ifaces);
1401 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1402 struct iface *iface;
1404 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1405 iface->ofp_port = OFPP_NONE;
1409 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1410 * already exist in the datapath and promote them to full fledged "struct
1411 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1412 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1413 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1414 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1415 garbage->ofp_port = ofproto_port.ofp_port;
1416 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1420 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1421 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1422 * later they can be added to ofproto. */
1423 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1424 struct iface *iface, *iface_next;
1426 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1427 if (iface->ofp_port == OFPP_NONE) {
1428 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1429 iface_destroy(iface);
1433 if (list_is_empty(&port->ifaces)) {
1439 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1440 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1441 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1444 * If successful, returns 0 and stores the network device in '*netdevp'. On
1445 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1447 iface_do_create(const struct bridge *br,
1448 const struct if_cfg *if_cfg,
1449 ofp_port_t *ofp_portp, struct netdev **netdevp)
1451 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1452 const struct ovsrec_port *port_cfg = if_cfg->parent;
1453 struct netdev *netdev = NULL;
1456 if (netdev_is_reserved_name(iface_cfg->name)) {
1457 VLOG_WARN("could not create interface %s, name is reserved",
1463 error = netdev_open(iface_cfg->name,
1464 iface_get_type(iface_cfg, br->cfg), &netdev);
1466 VLOG_WARN("could not open network device %s (%s)",
1467 iface_cfg->name, strerror(error));
1471 error = iface_set_netdev_config(iface_cfg, netdev);
1476 if (*ofp_portp == OFPP_NONE) {
1477 ofp_port_t ofp_port = if_cfg->ofport;
1479 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1483 *ofp_portp = ofp_port;
1485 VLOG_INFO("bridge %s: added interface %s on port %d",
1486 br->name, iface_cfg->name, *ofp_portp);
1488 VLOG_DBG("bridge %s: interface %s is on port %d",
1489 br->name, iface_cfg->name, *ofp_portp);
1492 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1493 || iface_is_internal(iface_cfg, br->cfg)) {
1494 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1502 netdev_close(netdev);
1506 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1507 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1508 * automatically allocated for the iface. Takes ownership of and
1509 * deallocates 'if_cfg'.
1511 * Return true if an iface is successfully created, false otherwise. */
1513 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1515 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1516 const struct ovsrec_port *port_cfg = if_cfg->parent;
1518 struct netdev *netdev;
1519 struct iface *iface;
1524 /* Do the bits that can fail up front.
1526 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1527 * internal datastructures may not be consistent. Eventually, when port
1528 * additions and deletions are cheaper, these calls should be removed. */
1530 ovs_assert(!iface_lookup(br, iface_cfg->name));
1531 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1534 iface_set_ofport(iface_cfg, OFPP_NONE);
1535 iface_clear_db_record(iface_cfg);
1540 /* Get or create the port structure. */
1541 port = port_lookup(br, port_cfg->name);
1543 port = port_create(br, port_cfg);
1546 /* Create the iface structure. */
1547 iface = xzalloc(sizeof *iface);
1548 list_push_back(&port->ifaces, &iface->port_elem);
1549 hmap_insert(&br->iface_by_name, &iface->name_node,
1550 hash_string(iface_cfg->name, 0));
1552 iface->name = xstrdup(iface_cfg->name);
1553 iface->ofp_port = OFPP_NONE;
1554 iface->netdev = netdev;
1555 iface->type = iface_get_type(iface_cfg, br->cfg);
1556 iface->cfg = iface_cfg;
1558 iface_set_ofp_port(iface, ofp_port);
1560 /* Populate initial status in database. */
1561 iface_refresh_stats(iface);
1562 iface_refresh_status(iface);
1564 /* Add bond fake iface if necessary. */
1565 if (port_is_bond_fake_iface(port)) {
1566 struct ofproto_port ofproto_port;
1568 if (ofproto_port_query_by_name(br->ofproto, port->name,
1570 struct netdev *netdev;
1573 error = netdev_open(port->name, "internal", &netdev);
1575 ofp_port_t fake_ofp_port = if_cfg->ofport;
1577 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1578 netdev_close(netdev);
1580 VLOG_WARN("could not open network device %s (%s)",
1581 port->name, strerror(error));
1584 /* Already exists, nothing to do. */
1585 ofproto_port_destroy(&ofproto_port);
1590 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1596 /* Set forward BPDU option. */
1598 bridge_configure_forward_bpdu(struct bridge *br)
1600 ofproto_set_forward_bpdu(br->ofproto,
1601 smap_get_bool(&br->cfg->other_config,
1606 /* Set MAC learning table configuration for 'br'. */
1608 bridge_configure_mac_table(struct bridge *br)
1610 const char *idle_time_str;
1613 const char *mac_table_size_str;
1616 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1617 idle_time = (idle_time_str && atoi(idle_time_str)
1618 ? atoi(idle_time_str)
1619 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1621 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1622 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1623 ? atoi(mac_table_size_str)
1626 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1630 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1631 struct iface **hw_addr_iface)
1633 struct hmapx mirror_output_ports;
1636 bool found_addr = false;
1640 *hw_addr_iface = NULL;
1642 /* Did the user request a particular MAC? */
1643 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1644 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1645 if (eth_addr_is_multicast(ea)) {
1646 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1647 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1648 } else if (eth_addr_is_zero(ea)) {
1649 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1655 /* Mirror output ports don't participate in picking the local hardware
1656 * address. ofproto can't help us find out whether a given port is a
1657 * mirror output because we haven't configured mirrors yet, so we need to
1658 * accumulate them ourselves. */
1659 hmapx_init(&mirror_output_ports);
1660 for (i = 0; i < br->cfg->n_mirrors; i++) {
1661 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1662 if (m->output_port) {
1663 hmapx_add(&mirror_output_ports, m->output_port);
1667 /* Otherwise choose the minimum non-local MAC address among all of the
1669 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1670 uint8_t iface_ea[ETH_ADDR_LEN];
1671 struct iface *candidate;
1672 struct iface *iface;
1674 /* Mirror output ports don't participate. */
1675 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1679 /* Choose the MAC address to represent the port. */
1681 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1682 /* Find the interface with this Ethernet address (if any) so that
1683 * we can provide the correct devname to the caller. */
1684 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1685 uint8_t candidate_ea[ETH_ADDR_LEN];
1686 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1687 && eth_addr_equals(iface_ea, candidate_ea)) {
1692 /* Choose the interface whose MAC address will represent the port.
1693 * The Linux kernel bonding code always chooses the MAC address of
1694 * the first slave added to a bond, and the Fedora networking
1695 * scripts always add slaves to a bond in alphabetical order, so
1696 * for compatibility we choose the interface with the name that is
1697 * first in alphabetical order. */
1698 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1699 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1704 /* The local port doesn't count (since we're trying to choose its
1705 * MAC address anyway). */
1706 if (iface->ofp_port == OFPP_LOCAL) {
1711 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1717 /* Compare against our current choice. */
1718 if (!eth_addr_is_multicast(iface_ea) &&
1719 !eth_addr_is_local(iface_ea) &&
1720 !eth_addr_is_reserved(iface_ea) &&
1721 !eth_addr_is_zero(iface_ea) &&
1722 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1724 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1725 *hw_addr_iface = iface;
1731 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1732 *hw_addr_iface = NULL;
1735 hmapx_destroy(&mirror_output_ports);
1738 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1739 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1740 * an interface on 'br', then that interface must be passed in as
1741 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1742 * 'hw_addr_iface' must be passed in as a null pointer. */
1744 bridge_pick_datapath_id(struct bridge *br,
1745 const uint8_t bridge_ea[ETH_ADDR_LEN],
1746 struct iface *hw_addr_iface)
1749 * The procedure for choosing a bridge MAC address will, in the most
1750 * ordinary case, also choose a unique MAC that we can use as a datapath
1751 * ID. In some special cases, though, multiple bridges will end up with
1752 * the same MAC address. This is OK for the bridges, but it will confuse
1753 * the OpenFlow controller, because each datapath needs a unique datapath
1756 * Datapath IDs must be unique. It is also very desirable that they be
1757 * stable from one run to the next, so that policy set on a datapath
1760 const char *datapath_id;
1763 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1764 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1768 if (!hw_addr_iface) {
1770 * A purely internal bridge, that is, one that has no non-virtual
1771 * network devices on it at all, is difficult because it has no
1772 * natural unique identifier at all.
1774 * When the host is a XenServer, we handle this case by hashing the
1775 * host's UUID with the name of the bridge. Names of bridges are
1776 * persistent across XenServer reboots, although they can be reused if
1777 * an internal network is destroyed and then a new one is later
1778 * created, so this is fairly effective.
1780 * When the host is not a XenServer, we punt by using a random MAC
1781 * address on each run.
1783 const char *host_uuid = xenserver_get_host_uuid();
1785 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1786 dpid = dpid_from_hash(combined, strlen(combined));
1792 return eth_addr_to_uint64(bridge_ea);
1796 dpid_from_hash(const void *data, size_t n)
1798 uint8_t hash[SHA1_DIGEST_SIZE];
1800 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1801 sha1_bytes(data, n, hash);
1802 eth_addr_mark_random(hash);
1803 return eth_addr_to_uint64(hash);
1807 iface_refresh_status(struct iface *iface)
1811 enum netdev_features current;
1815 uint8_t mac[ETH_ADDR_LEN];
1818 if (iface_is_synthetic(iface)) {
1824 if (!netdev_get_status(iface->netdev, &smap)) {
1825 ovsrec_interface_set_status(iface->cfg, &smap);
1827 ovsrec_interface_set_status(iface->cfg, NULL);
1830 smap_destroy(&smap);
1832 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1833 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1835 ovsrec_interface_set_duplex(iface->cfg,
1836 netdev_features_is_full_duplex(current)
1838 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1840 ovsrec_interface_set_duplex(iface->cfg, NULL);
1841 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1844 error = netdev_get_mtu(iface->netdev, &mtu);
1847 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1849 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1852 error = netdev_get_etheraddr(iface->netdev, mac);
1854 char mac_string[32];
1856 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1857 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1859 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1863 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1866 iface_refresh_cfm_stats(struct iface *iface)
1868 const struct ovsrec_interface *cfg = iface->cfg;
1869 struct ofproto_cfm_status status;
1871 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1872 iface->ofp_port, &status)) {
1873 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1874 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1875 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1876 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1877 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1879 const char *reasons[CFM_FAULT_N_REASONS];
1880 int64_t cfm_health = status.health;
1881 bool faulted = status.faults != 0;
1884 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1887 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1888 int reason = 1 << i;
1889 if (status.faults & reason) {
1890 reasons[j++] = cfm_fault_reason_to_str(reason);
1893 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1895 if (status.remote_opstate >= 0) {
1896 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1897 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1899 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1902 ovsrec_interface_set_cfm_remote_mpids(cfg,
1903 (const int64_t *)status.rmps,
1905 if (cfm_health >= 0) {
1906 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1908 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1914 iface_refresh_stats(struct iface *iface)
1916 #define IFACE_STATS \
1917 IFACE_STAT(rx_packets, "rx_packets") \
1918 IFACE_STAT(tx_packets, "tx_packets") \
1919 IFACE_STAT(rx_bytes, "rx_bytes") \
1920 IFACE_STAT(tx_bytes, "tx_bytes") \
1921 IFACE_STAT(rx_dropped, "rx_dropped") \
1922 IFACE_STAT(tx_dropped, "tx_dropped") \
1923 IFACE_STAT(rx_errors, "rx_errors") \
1924 IFACE_STAT(tx_errors, "tx_errors") \
1925 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1926 IFACE_STAT(rx_over_errors, "rx_over_err") \
1927 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1928 IFACE_STAT(collisions, "collisions")
1930 #define IFACE_STAT(MEMBER, NAME) + 1
1931 enum { N_IFACE_STATS = IFACE_STATS };
1933 int64_t values[N_IFACE_STATS];
1934 char *keys[N_IFACE_STATS];
1937 struct netdev_stats stats;
1939 if (iface_is_synthetic(iface)) {
1943 /* Intentionally ignore return value, since errors will set 'stats' to
1944 * all-1s, and we will deal with that correctly below. */
1945 netdev_get_stats(iface->netdev, &stats);
1947 /* Copy statistics into keys[] and values[]. */
1949 #define IFACE_STAT(MEMBER, NAME) \
1950 if (stats.MEMBER != UINT64_MAX) { \
1952 values[n] = stats.MEMBER; \
1957 ovs_assert(n <= N_IFACE_STATS);
1959 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1964 br_refresh_stp_status(struct bridge *br)
1966 struct smap smap = SMAP_INITIALIZER(&smap);
1967 struct ofproto *ofproto = br->ofproto;
1968 struct ofproto_stp_status status;
1970 if (ofproto_get_stp_status(ofproto, &status)) {
1974 if (!status.enabled) {
1975 ovsrec_bridge_set_status(br->cfg, NULL);
1979 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1980 STP_ID_ARGS(status.bridge_id));
1981 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1982 STP_ID_ARGS(status.designated_root));
1983 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1985 ovsrec_bridge_set_status(br->cfg, &smap);
1986 smap_destroy(&smap);
1990 port_refresh_stp_status(struct port *port)
1992 struct ofproto *ofproto = port->bridge->ofproto;
1993 struct iface *iface;
1994 struct ofproto_port_stp_status status;
1996 int64_t int_values[3];
1999 if (port_is_synthetic(port)) {
2003 /* STP doesn't currently support bonds. */
2004 if (!list_is_singleton(&port->ifaces)) {
2005 ovsrec_port_set_status(port->cfg, NULL);
2009 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2011 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2015 if (!status.enabled) {
2016 ovsrec_port_set_status(port->cfg, NULL);
2017 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2021 /* Set Status column. */
2023 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2024 smap_add(&smap, "stp_state", stp_state_name(status.state));
2025 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2026 smap_add(&smap, "stp_role", stp_role_name(status.role));
2027 ovsrec_port_set_status(port->cfg, &smap);
2028 smap_destroy(&smap);
2030 /* Set Statistics column. */
2031 keys[0] = "stp_tx_count";
2032 int_values[0] = status.tx_count;
2033 keys[1] = "stp_rx_count";
2034 int_values[1] = status.rx_count;
2035 keys[2] = "stp_error_count";
2036 int_values[2] = status.error_count;
2038 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2039 ARRAY_SIZE(int_values));
2043 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2045 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2049 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2051 bool enable = enable_system_stats(cfg);
2053 system_stats_enable(enable);
2055 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2060 run_system_stats(void)
2062 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2065 stats = system_stats_run();
2067 struct ovsdb_idl_txn *txn;
2068 struct ovsdb_datum datum;
2070 txn = ovsdb_idl_txn_create(idl);
2071 ovsdb_datum_from_smap(&datum, stats);
2072 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2074 ovsdb_idl_txn_commit(txn);
2075 ovsdb_idl_txn_destroy(txn);
2081 static inline const char *
2082 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2085 case OFPCR12_ROLE_EQUAL:
2087 case OFPCR12_ROLE_MASTER:
2089 case OFPCR12_ROLE_SLAVE:
2091 case OFPCR12_ROLE_NOCHANGE:
2093 return "*** INVALID ROLE ***";
2098 refresh_controller_status(void)
2102 const struct ovsrec_controller *cfg;
2106 /* Accumulate status for controllers on all bridges. */
2107 HMAP_FOR_EACH (br, node, &all_bridges) {
2108 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2111 /* Update each controller in the database with current status. */
2112 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2113 struct ofproto_controller_info *cinfo =
2114 shash_find_data(&info, cfg->target);
2117 struct smap smap = SMAP_INITIALIZER(&smap);
2118 const char **values = cinfo->pairs.values;
2119 const char **keys = cinfo->pairs.keys;
2122 for (i = 0; i < cinfo->pairs.n; i++) {
2123 smap_add(&smap, keys[i], values[i]);
2126 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2127 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2129 ovsrec_controller_set_status(cfg, &smap);
2130 smap_destroy(&smap);
2132 ovsrec_controller_set_is_connected(cfg, false);
2133 ovsrec_controller_set_role(cfg, NULL);
2134 ovsrec_controller_set_status(cfg, NULL);
2138 ofproto_free_ofproto_controller_info(&info);
2143 * Some information in the database must be kept as up-to-date as possible to
2144 * allow controllers to respond rapidly to network outages. We call these
2145 * statistics "instant" stats.
2147 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2148 * assuming that they've changed. The only means we have to determine whether
2149 * they have changed are:
2151 * - Try to commit changes to the database. If nothing changed, then
2152 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2155 * - instant_stats_run() is called late in the run loop, after anything that
2156 * might change any of the instant stats.
2158 * We use these two facts together to avoid waking the process up every
2159 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2162 /* Minimum interval between writing updates to the instant stats to the
2164 #define INSTANT_INTERVAL_MSEC 100
2166 /* Current instant stats database transaction, NULL if there is no ongoing
2168 static struct ovsdb_idl_txn *instant_txn;
2170 /* Next time (in msec on monotonic clock) at which we will update the instant
2172 static long long int instant_next_txn = LLONG_MIN;
2174 /* True if the run loop has run since we last saw that the instant stats were
2175 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2176 * to refresh the instant stats. */
2177 static bool instant_stats_could_have_changed;
2180 instant_stats_run(void)
2182 enum ovsdb_idl_txn_status status;
2184 instant_stats_could_have_changed = true;
2189 if (time_msec() < instant_next_txn) {
2192 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2194 instant_txn = ovsdb_idl_txn_create(idl);
2195 HMAP_FOR_EACH (br, node, &all_bridges) {
2196 struct iface *iface;
2199 br_refresh_stp_status(br);
2201 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2202 port_refresh_stp_status(port);
2205 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2206 enum netdev_flags flags;
2208 const char *link_state;
2209 int64_t link_resets;
2212 if (iface_is_synthetic(iface)) {
2216 current = ofproto_port_is_lacp_current(br->ofproto,
2220 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2222 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2225 error = netdev_get_flags(iface->netdev, &flags);
2227 const char *state = flags & NETDEV_UP ? "up" : "down";
2228 ovsrec_interface_set_admin_state(iface->cfg, state);
2230 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2233 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2234 ovsrec_interface_set_link_state(iface->cfg, link_state);
2236 link_resets = netdev_get_carrier_resets(iface->netdev);
2237 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2239 iface_refresh_cfm_stats(iface);
2242 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2244 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2245 smap_destroy(&smap);
2251 status = ovsdb_idl_txn_commit(instant_txn);
2252 if (status != TXN_INCOMPLETE) {
2253 ovsdb_idl_txn_destroy(instant_txn);
2256 if (status == TXN_UNCHANGED) {
2257 instant_stats_could_have_changed = false;
2262 instant_stats_wait(void)
2265 ovsdb_idl_txn_wait(instant_txn);
2266 } else if (instant_stats_could_have_changed) {
2267 poll_timer_wait_until(instant_next_txn);
2271 /* Performs periodic activity required by bridges that needs to be done with
2272 * the least possible latency.
2274 * It makes sense to call this function a couple of times per poll loop, to
2275 * provide a significant performance boost on some benchmarks with ofprotos
2276 * that use the ofproto-dpif implementation. */
2278 bridge_run_fast(void)
2285 ofproto_enumerate_types(&types);
2286 SSET_FOR_EACH (type, &types) {
2287 ofproto_type_run_fast(type);
2289 sset_destroy(&types);
2291 HMAP_FOR_EACH (br, node, &all_bridges) {
2292 ofproto_run_fast(br->ofproto);
2299 static struct ovsrec_open_vswitch null_cfg;
2300 const struct ovsrec_open_vswitch *cfg;
2301 struct ovsdb_idl_txn *reconf_txn = NULL;
2305 bool vlan_splinters_changed;
2308 ovsrec_open_vswitch_init(&null_cfg);
2310 /* (Re)configure if necessary. */
2311 if (!reconfiguring) {
2314 if (ovsdb_idl_is_lock_contended(idl)) {
2315 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2316 struct bridge *br, *next_br;
2318 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2319 "disabling this process (pid %ld) until it goes away",
2320 (long int) getpid());
2322 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2325 /* Since we will not be running system_stats_run() in this process
2326 * with the current situation of multiple ovs-vswitchd daemons,
2327 * disable system stats collection. */
2328 system_stats_enable(false);
2330 } else if (!ovsdb_idl_has_lock(idl)) {
2334 cfg = ovsrec_open_vswitch_first(idl);
2336 /* Initialize the ofproto library. This only needs to run once, but
2337 * it must be done after the configuration is set. If the
2338 * initialization has already occurred, bridge_init_ofproto()
2339 * returns immediately. */
2340 bridge_init_ofproto(cfg);
2342 /* Once the value of flow-restore-wait is false, we no longer should
2343 * check its value from the database. */
2344 if (cfg && ofproto_get_flow_restore_wait()) {
2345 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2346 "flow-restore-wait", false));
2349 /* Let each datapath type do the work that it needs to do. */
2351 ofproto_enumerate_types(&types);
2352 SSET_FOR_EACH (type, &types) {
2353 ofproto_type_run(type);
2355 sset_destroy(&types);
2357 /* Let each bridge do the work that it needs to do. */
2358 HMAP_FOR_EACH (br, node, &all_bridges) {
2359 ofproto_run(br->ofproto);
2362 /* Re-configure SSL. We do this on every trip through the main loop,
2363 * instead of just when the database changes, because the contents of the
2364 * key and certificate files can change without the database changing.
2366 * We do this before bridge_reconfigure() because that function might
2367 * initiate SSL connections and thus requires SSL to be configured. */
2368 if (cfg && cfg->ssl) {
2369 const struct ovsrec_ssl *ssl = cfg->ssl;
2371 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2372 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2375 if (!reconfiguring) {
2376 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2377 * usage has changed. */
2378 vlan_splinters_changed = false;
2379 if (vlan_splinters_enabled_anywhere) {
2380 HMAP_FOR_EACH (br, node, &all_bridges) {
2381 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2382 vlan_splinters_changed = true;
2388 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2389 idl_seqno = ovsdb_idl_get_seqno(idl);
2391 reconf_txn = ovsdb_idl_txn_create(idl);
2392 bridge_reconfigure(cfg);
2394 /* We still need to reconfigure to avoid dangling pointers to
2395 * now-destroyed ovsrec structures inside bridge data. */
2396 bridge_reconfigure(&null_cfg);
2401 if (reconfiguring) {
2403 reconf_txn = ovsdb_idl_txn_create(idl);
2406 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2407 reconfiguring = false;
2410 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2413 /* If we are completing our initial configuration for this run
2414 * of ovs-vswitchd, then keep the transaction around to monitor
2415 * it for completion. */
2416 if (!initial_config_done) {
2417 initial_config_done = true;
2418 daemonize_txn = reconf_txn;
2425 ovsdb_idl_txn_commit(reconf_txn);
2426 ovsdb_idl_txn_destroy(reconf_txn);
2430 if (daemonize_txn) {
2431 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2432 if (status != TXN_INCOMPLETE) {
2433 ovsdb_idl_txn_destroy(daemonize_txn);
2434 daemonize_txn = NULL;
2436 /* ovs-vswitchd has completed initialization, so allow the
2437 * process that forked us to exit successfully. */
2438 daemonize_complete();
2440 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2444 /* Refresh interface and mirror stats if necessary. */
2445 if (time_msec() >= iface_stats_timer) {
2447 struct ovsdb_idl_txn *txn;
2449 txn = ovsdb_idl_txn_create(idl);
2450 HMAP_FOR_EACH (br, node, &all_bridges) {
2454 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2455 struct iface *iface;
2457 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2458 iface_refresh_stats(iface);
2459 iface_refresh_status(iface);
2463 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2464 mirror_refresh_stats(m);
2468 refresh_controller_status();
2469 ovsdb_idl_txn_commit(txn);
2470 ovsdb_idl_txn_destroy(txn); /* XXX */
2473 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2477 instant_stats_run();
2486 ovsdb_idl_wait(idl);
2487 if (daemonize_txn) {
2488 ovsdb_idl_txn_wait(daemonize_txn);
2491 if (reconfiguring) {
2492 poll_immediate_wake();
2496 ofproto_enumerate_types(&types);
2497 SSET_FOR_EACH (type, &types) {
2498 ofproto_type_wait(type);
2500 sset_destroy(&types);
2502 if (!hmap_is_empty(&all_bridges)) {
2505 HMAP_FOR_EACH (br, node, &all_bridges) {
2506 ofproto_wait(br->ofproto);
2508 poll_timer_wait_until(iface_stats_timer);
2511 system_stats_wait();
2512 instant_stats_wait();
2515 /* Adds some memory usage statistics for bridges into 'usage', for use with
2516 * memory_report(). */
2518 bridge_get_memory_usage(struct simap *usage)
2522 HMAP_FOR_EACH (br, node, &all_bridges) {
2523 ofproto_get_memory_usage(br->ofproto, usage);
2527 /* QoS unixctl user interface functions. */
2529 struct qos_unixctl_show_cbdata {
2531 struct iface *iface;
2535 qos_unixctl_show_cb(unsigned int queue_id,
2536 const struct smap *details,
2539 struct qos_unixctl_show_cbdata *data = aux;
2540 struct ds *ds = data->ds;
2541 struct iface *iface = data->iface;
2542 struct netdev_queue_stats stats;
2543 struct smap_node *node;
2546 ds_put_cstr(ds, "\n");
2548 ds_put_format(ds, "Queue %u:\n", queue_id);
2550 ds_put_cstr(ds, "Default:\n");
2553 SMAP_FOR_EACH (node, details) {
2554 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2557 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2559 if (stats.tx_packets != UINT64_MAX) {
2560 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2563 if (stats.tx_bytes != UINT64_MAX) {
2564 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2567 if (stats.tx_errors != UINT64_MAX) {
2568 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2571 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2572 queue_id, strerror(error));
2577 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2578 const char *argv[], void *aux OVS_UNUSED)
2580 struct ds ds = DS_EMPTY_INITIALIZER;
2581 struct smap smap = SMAP_INITIALIZER(&smap);
2582 struct iface *iface;
2584 struct smap_node *node;
2585 struct qos_unixctl_show_cbdata data;
2588 iface = iface_find(argv[1]);
2590 unixctl_command_reply_error(conn, "no such interface");
2594 netdev_get_qos(iface->netdev, &type, &smap);
2596 if (*type != '\0') {
2597 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2599 SMAP_FOR_EACH (node, &smap) {
2600 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2605 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2608 ds_put_format(&ds, "failed to dump queues: %s", strerror(error));
2610 unixctl_command_reply(conn, ds_cstr(&ds));
2612 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2613 unixctl_command_reply_error(conn, ds_cstr(&ds));
2616 smap_destroy(&smap);
2620 /* Bridge reconfiguration functions. */
2622 bridge_create(const struct ovsrec_bridge *br_cfg)
2626 ovs_assert(!bridge_lookup(br_cfg->name));
2627 br = xzalloc(sizeof *br);
2629 br->name = xstrdup(br_cfg->name);
2630 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2633 /* Derive the default Ethernet address from the bridge's UUID. This should
2634 * be unique and it will be stable between ovs-vswitchd runs. */
2635 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2636 eth_addr_mark_random(br->default_ea);
2638 hmap_init(&br->ports);
2639 hmap_init(&br->ifaces);
2640 hmap_init(&br->iface_by_name);
2641 hmap_init(&br->mirrors);
2643 hmap_init(&br->if_cfg_todo);
2644 list_init(&br->ofpp_garbage);
2646 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2650 bridge_destroy(struct bridge *br)
2653 struct mirror *mirror, *next_mirror;
2654 struct port *port, *next_port;
2655 struct if_cfg *if_cfg, *next_if_cfg;
2656 struct ofpp_garbage *garbage, *next_garbage;
2658 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2661 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2662 mirror_destroy(mirror);
2664 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2665 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2668 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2669 &br->ofpp_garbage) {
2670 list_remove(&garbage->list_node);
2674 hmap_remove(&all_bridges, &br->node);
2675 ofproto_destroy(br->ofproto);
2676 hmap_destroy(&br->ifaces);
2677 hmap_destroy(&br->ports);
2678 hmap_destroy(&br->iface_by_name);
2679 hmap_destroy(&br->mirrors);
2680 hmap_destroy(&br->if_cfg_todo);
2687 static struct bridge *
2688 bridge_lookup(const char *name)
2692 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2693 if (!strcmp(br->name, name)) {
2700 /* Handle requests for a listing of all flows known by the OpenFlow
2701 * stack, including those normally hidden. */
2703 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2704 const char *argv[], void *aux OVS_UNUSED)
2709 br = bridge_lookup(argv[1]);
2711 unixctl_command_reply_error(conn, "Unknown bridge");
2716 ofproto_get_all_flows(br->ofproto, &results);
2718 unixctl_command_reply(conn, ds_cstr(&results));
2719 ds_destroy(&results);
2722 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2723 * connections and reconnect. If BRIDGE is not specified, then all bridges
2724 * drop their controller connections and reconnect. */
2726 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2727 const char *argv[], void *aux OVS_UNUSED)
2731 br = bridge_lookup(argv[1]);
2733 unixctl_command_reply_error(conn, "Unknown bridge");
2736 ofproto_reconnect_controllers(br->ofproto);
2738 HMAP_FOR_EACH (br, node, &all_bridges) {
2739 ofproto_reconnect_controllers(br->ofproto);
2742 unixctl_command_reply(conn, NULL);
2746 bridge_get_controllers(const struct bridge *br,
2747 struct ovsrec_controller ***controllersp)
2749 struct ovsrec_controller **controllers;
2750 size_t n_controllers;
2752 controllers = br->cfg->controller;
2753 n_controllers = br->cfg->n_controller;
2755 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2761 *controllersp = controllers;
2763 return n_controllers;
2767 bridge_queue_if_cfg(struct bridge *br,
2768 const struct ovsrec_interface *cfg,
2769 const struct ovsrec_port *parent)
2771 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2774 if_cfg->parent = parent;
2775 if_cfg->ofport = iface_pick_ofport(cfg);
2776 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2777 hash_string(if_cfg->cfg->name, 0));
2780 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2781 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2782 * 'br' needs to complete its configuration. */
2784 bridge_add_del_ports(struct bridge *br,
2785 const unsigned long int *splinter_vlans)
2787 struct shash_node *port_node;
2788 struct port *port, *next;
2789 struct shash new_ports;
2792 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2794 /* Collect new ports. */
2795 shash_init(&new_ports);
2796 for (i = 0; i < br->cfg->n_ports; i++) {
2797 const char *name = br->cfg->ports[i]->name;
2798 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2799 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2803 if (bridge_get_controllers(br, NULL)
2804 && !shash_find(&new_ports, br->name)) {
2805 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2806 br->name, br->name);
2808 ovsrec_interface_init(&br->synth_local_iface);
2809 ovsrec_port_init(&br->synth_local_port);
2811 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2812 br->synth_local_port.n_interfaces = 1;
2813 br->synth_local_port.name = br->name;
2815 br->synth_local_iface.name = br->name;
2816 br->synth_local_iface.type = "internal";
2818 br->synth_local_ifacep = &br->synth_local_iface;
2820 shash_add(&new_ports, br->name, &br->synth_local_port);
2823 if (splinter_vlans) {
2824 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2827 /* Get rid of deleted ports.
2828 * Get rid of deleted interfaces on ports that still exist. */
2829 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2830 port->cfg = shash_find_data(&new_ports, port->name);
2834 port_del_ifaces(port);
2838 /* Update iface->cfg and iface->type in interfaces that still exist.
2839 * Add new interfaces to creation queue. */
2840 SHASH_FOR_EACH (port_node, &new_ports) {
2841 const struct ovsrec_port *port = port_node->data;
2844 for (i = 0; i < port->n_interfaces; i++) {
2845 const struct ovsrec_interface *cfg = port->interfaces[i];
2846 struct iface *iface = iface_lookup(br, cfg->name);
2847 const char *type = iface_get_type(cfg, br->cfg);
2852 } else if (!strcmp(type, "null")) {
2853 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2854 " may be removed in February 2013. Please email"
2855 " dev@openvswitch.org with concerns.",
2858 bridge_queue_if_cfg(br, cfg, port);
2863 shash_destroy(&new_ports);
2866 /* Initializes 'oc' appropriately as a management service controller for
2869 * The caller must free oc->target when it is no longer needed. */
2871 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2872 struct ofproto_controller *oc)
2874 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2875 oc->max_backoff = 0;
2876 oc->probe_interval = 60;
2877 oc->band = OFPROTO_OUT_OF_BAND;
2879 oc->burst_limit = 0;
2880 oc->enable_async_msgs = true;
2883 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2885 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2886 struct ofproto_controller *oc)
2890 oc->target = c->target;
2891 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2892 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2893 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2894 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2895 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2896 oc->burst_limit = (c->controller_burst_limit
2897 ? *c->controller_burst_limit : 0);
2898 oc->enable_async_msgs = (!c->enable_async_messages
2899 || *c->enable_async_messages);
2900 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2901 if (dscp < 0 || dscp > 63) {
2902 dscp = DSCP_DEFAULT;
2907 /* Configures the IP stack for 'br''s local interface properly according to the
2908 * configuration in 'c'. */
2910 bridge_configure_local_iface_netdev(struct bridge *br,
2911 struct ovsrec_controller *c)
2913 struct netdev *netdev;
2914 struct in_addr mask, gateway;
2916 struct iface *local_iface;
2919 /* If there's no local interface or no IP address, give up. */
2920 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2921 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2925 /* Bring up the local interface. */
2926 netdev = local_iface->netdev;
2927 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2929 /* Configure the IP address and netmask. */
2930 if (!c->local_netmask
2931 || !inet_aton(c->local_netmask, &mask)
2933 mask.s_addr = guess_netmask(ip.s_addr);
2935 if (!netdev_set_in4(netdev, ip, mask)) {
2936 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2937 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2940 /* Configure the default gateway. */
2941 if (c->local_gateway
2942 && inet_aton(c->local_gateway, &gateway)
2943 && gateway.s_addr) {
2944 if (!netdev_add_router(netdev, gateway)) {
2945 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2946 br->name, IP_ARGS(gateway.s_addr));
2951 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2952 * in either string are treated as equal to any number of slashes in the other,
2953 * e.g. "x///y" is equal to "x/y".
2955 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2956 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2957 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2958 * 'b' against a prefix of 'a'.
2961 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2963 const char *b_start = b;
2965 if (b - b_start >= b_stoplen) {
2967 } else if (*a != *b) {
2969 } else if (*a == '/') {
2970 a += strspn(a, "/");
2971 b += strspn(b, "/");
2972 } else if (*a == '\0') {
2982 bridge_configure_remotes(struct bridge *br,
2983 const struct sockaddr_in *managers, size_t n_managers)
2985 bool disable_in_band;
2987 struct ovsrec_controller **controllers;
2988 size_t n_controllers;
2990 enum ofproto_fail_mode fail_mode;
2992 struct ofproto_controller *ocs;
2996 /* Check if we should disable in-band control on this bridge. */
2997 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3000 /* Set OpenFlow queue ID for in-band control. */
3001 ofproto_set_in_band_queue(br->ofproto,
3002 smap_get_int(&br->cfg->other_config,
3003 "in-band-queue", -1));
3005 if (disable_in_band) {
3006 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3008 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3011 n_controllers = bridge_get_controllers(br, &controllers);
3013 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3016 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3017 for (i = 0; i < n_controllers; i++) {
3018 struct ovsrec_controller *c = controllers[i];
3020 if (!strncmp(c->target, "punix:", 6)
3021 || !strncmp(c->target, "unix:", 5)) {
3022 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3025 if (!strncmp(c->target, "unix:", 5)) {
3026 /* Connect to a listening socket */
3027 whitelist = xasprintf("unix:%s/", ovs_rundir());
3028 if (strchr(c->target, '/') &&
3029 !equal_pathnames(c->target, whitelist,
3030 strlen(whitelist))) {
3031 /* Absolute path specified, but not in ovs_rundir */
3032 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3033 "controller \"%s\" due to possibility for "
3034 "remote exploit. Instead, specify socket "
3035 "in whitelisted \"%s\" or connect to "
3036 "\"unix:%s/%s.mgmt\" (which is always "
3037 "available without special configuration).",
3038 br->name, c->target, whitelist,
3039 ovs_rundir(), br->name);
3044 whitelist = xasprintf("punix:%s/%s.controller",
3045 ovs_rundir(), br->name);
3046 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3047 /* Prevent remote ovsdb-server users from accessing
3048 * arbitrary Unix domain sockets and overwriting arbitrary
3050 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3051 "controller \"%s\" due to possibility of "
3052 "overwriting local files. Instead, specify "
3053 "whitelisted \"%s\" or connect to "
3054 "\"unix:%s/%s.mgmt\" (which is always "
3055 "available without special configuration).",
3056 br->name, c->target, whitelist,
3057 ovs_rundir(), br->name);
3066 bridge_configure_local_iface_netdev(br, c);
3067 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3068 if (disable_in_band) {
3069 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3074 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3075 bridge_get_allowed_versions(br));
3076 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3079 /* Set the fail-mode. */
3080 fail_mode = !br->cfg->fail_mode
3081 || !strcmp(br->cfg->fail_mode, "standalone")
3082 ? OFPROTO_FAIL_STANDALONE
3083 : OFPROTO_FAIL_SECURE;
3084 ofproto_set_fail_mode(br->ofproto, fail_mode);
3086 /* Configure OpenFlow controller connection snooping. */
3087 if (!ofproto_has_snoops(br->ofproto)) {
3091 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3092 ovs_rundir(), br->name));
3093 ofproto_set_snoops(br->ofproto, &snoops);
3094 sset_destroy(&snoops);
3099 bridge_configure_tables(struct bridge *br)
3101 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3105 n_tables = ofproto_get_n_tables(br->ofproto);
3107 for (i = 0; i < n_tables; i++) {
3108 struct ofproto_table_settings s;
3111 s.max_flows = UINT_MAX;
3115 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3116 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3119 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3120 s.max_flows = *cfg->flow_limit;
3122 if (cfg->overflow_policy
3123 && !strcmp(cfg->overflow_policy, "evict")) {
3126 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3127 for (k = 0; k < cfg->n_groups; k++) {
3128 const char *string = cfg->groups[k];
3131 msg = mf_parse_subfield__(&s.groups[k], &string);
3133 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3134 "'groups' (%s)", br->name, i, msg);
3136 } else if (*string) {
3137 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3138 "element '%s' contains trailing garbage",
3139 br->name, i, cfg->groups[k]);
3147 ofproto_configure_table(br->ofproto, i, &s);
3151 for (; j < br->cfg->n_flow_tables; j++) {
3152 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3153 "%"PRId64" not supported by this datapath", br->name,
3154 br->cfg->key_flow_tables[j]);
3159 bridge_configure_dp_desc(struct bridge *br)
3161 ofproto_set_dp_desc(br->ofproto,
3162 smap_get(&br->cfg->other_config, "dp-desc"));
3165 /* Port functions. */
3167 static struct port *
3168 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3172 port = xzalloc(sizeof *port);
3174 port->name = xstrdup(cfg->name);
3176 list_init(&port->ifaces);
3178 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3182 /* Deletes interfaces from 'port' that are no longer configured for it. */
3184 port_del_ifaces(struct port *port)
3186 struct iface *iface, *next;
3187 struct sset new_ifaces;
3190 /* Collect list of new interfaces. */
3191 sset_init(&new_ifaces);
3192 for (i = 0; i < port->cfg->n_interfaces; i++) {
3193 const char *name = port->cfg->interfaces[i]->name;
3194 const char *type = port->cfg->interfaces[i]->type;
3195 if (strcmp(type, "null")) {
3196 sset_add(&new_ifaces, name);
3200 /* Get rid of deleted interfaces. */
3201 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3202 if (!sset_contains(&new_ifaces, iface->name)) {
3203 iface_destroy(iface);
3207 sset_destroy(&new_ifaces);
3211 port_destroy(struct port *port)
3214 struct bridge *br = port->bridge;
3215 struct iface *iface, *next;
3218 ofproto_bundle_unregister(br->ofproto, port);
3221 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3222 iface_destroy(iface);
3225 hmap_remove(&br->ports, &port->hmap_node);
3231 static struct port *
3232 port_lookup(const struct bridge *br, const char *name)
3236 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3238 if (!strcmp(port->name, name)) {
3246 enable_lacp(struct port *port, bool *activep)
3248 if (!port->cfg->lacp) {
3249 /* XXX when LACP implementation has been sufficiently tested, enable by
3250 * default and make active on bonded ports. */
3252 } else if (!strcmp(port->cfg->lacp, "off")) {
3254 } else if (!strcmp(port->cfg->lacp, "active")) {
3257 } else if (!strcmp(port->cfg->lacp, "passive")) {
3261 VLOG_WARN("port %s: unknown LACP mode %s",
3262 port->name, port->cfg->lacp);
3267 static struct lacp_settings *
3268 port_configure_lacp(struct port *port, struct lacp_settings *s)
3270 const char *lacp_time, *system_id;
3273 if (!enable_lacp(port, &s->active)) {
3277 s->name = port->name;
3279 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3281 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3282 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3283 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3284 " address.", port->name, system_id);
3288 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3291 if (eth_addr_is_zero(s->id)) {
3292 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3296 /* Prefer bondable links if unspecified. */
3297 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3299 s->priority = (priority > 0 && priority <= UINT16_MAX
3301 : UINT16_MAX - !list_is_short(&port->ifaces));
3303 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3304 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3309 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3311 int priority, portid, key;
3313 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3314 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3316 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3318 if (portid <= 0 || portid > UINT16_MAX) {
3319 portid = ofp_to_u16(iface->ofp_port);
3322 if (priority <= 0 || priority > UINT16_MAX) {
3323 priority = UINT16_MAX;
3326 if (key < 0 || key > UINT16_MAX) {
3330 s->name = iface->name;
3332 s->priority = priority;
3337 port_configure_bond(struct port *port, struct bond_settings *s)
3339 const char *detect_s;
3340 struct iface *iface;
3341 int miimon_interval;
3343 s->name = port->name;
3345 if (port->cfg->bond_mode) {
3346 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3347 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3348 port->name, port->cfg->bond_mode,
3349 bond_mode_to_string(s->balance));
3352 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3354 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3355 * active-backup. At some point we should remove this warning. */
3356 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3357 " in previous versions, the default bond_mode was"
3358 " balance-slb", port->name,
3359 bond_mode_to_string(s->balance));
3361 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3362 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3363 "please use another bond type or disable flood_vlans",
3367 miimon_interval = smap_get_int(&port->cfg->other_config,
3368 "bond-miimon-interval", 0);
3369 if (miimon_interval <= 0) {
3370 miimon_interval = 200;
3373 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3374 if (!detect_s || !strcmp(detect_s, "carrier")) {
3375 miimon_interval = 0;
3376 } else if (strcmp(detect_s, "miimon")) {
3377 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3378 "defaulting to carrier", port->name, detect_s);
3379 miimon_interval = 0;
3382 s->up_delay = MAX(0, port->cfg->bond_updelay);
3383 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3384 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3385 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3386 "bond-rebalance-interval", 10000);
3387 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3388 s->rebalance_interval = 1000;
3391 s->fake_iface = port->cfg->bond_fake_iface;
3393 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3394 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3398 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3399 * instead of obtaining it from the database. */
3401 port_is_synthetic(const struct port *port)
3403 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3406 /* Interface functions. */
3409 iface_is_internal(const struct ovsrec_interface *iface,
3410 const struct ovsrec_bridge *br)
3412 /* The local port and "internal" ports are always "internal". */
3413 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3416 /* Returns the correct network device type for interface 'iface' in bridge
3419 iface_get_type(const struct ovsrec_interface *iface,
3420 const struct ovsrec_bridge *br)
3424 /* The local port always has type "internal". Other ports take
3425 * their type from the database and default to "system" if none is
3427 if (iface_is_internal(iface, br)) {
3430 type = iface->type[0] ? iface->type : "system";
3433 return ofproto_port_open_type(br->datapath_type, type);
3437 iface_destroy(struct iface *iface)
3440 struct port *port = iface->port;
3441 struct bridge *br = port->bridge;
3443 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3444 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3447 if (iface->ofp_port != OFPP_NONE) {
3448 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3451 list_remove(&iface->port_elem);
3452 hmap_remove(&br->iface_by_name, &iface->name_node);
3454 netdev_close(iface->netdev);
3461 static struct iface *
3462 iface_lookup(const struct bridge *br, const char *name)
3464 struct iface *iface;
3466 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3467 &br->iface_by_name) {
3468 if (!strcmp(iface->name, name)) {
3476 static struct iface *
3477 iface_find(const char *name)
3479 const struct bridge *br;
3481 HMAP_FOR_EACH (br, node, &all_bridges) {
3482 struct iface *iface = iface_lookup(br, name);
3491 static struct if_cfg *
3492 if_cfg_lookup(const struct bridge *br, const char *name)
3494 struct if_cfg *if_cfg;
3496 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3498 if (!strcmp(if_cfg->cfg->name, name)) {
3506 static struct iface *
3507 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3509 struct iface *iface;
3511 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3513 if (iface->ofp_port == ofp_port) {
3520 /* Set Ethernet address of 'iface', if one is specified in the configuration
3523 iface_set_mac(struct iface *iface)
3525 uint8_t ea[ETH_ADDR_LEN];
3527 if (!strcmp(iface->type, "internal")
3528 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3529 if (iface->ofp_port == OFPP_LOCAL) {
3530 VLOG_ERR("interface %s: ignoring mac in Interface record "
3531 "(use Bridge record to set local port's mac)",
3533 } else if (eth_addr_is_multicast(ea)) {
3534 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3537 int error = netdev_set_etheraddr(iface->netdev, ea);
3539 VLOG_ERR("interface %s: setting MAC failed (%s)",
3540 iface->name, strerror(error));
3546 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3548 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3551 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3552 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3553 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3557 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3558 * sets the "ofport" field to -1.
3560 * This is appropriate when 'if_cfg''s interface cannot be created or is
3561 * otherwise invalid. */
3563 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3565 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3566 ovsrec_interface_set_status(if_cfg, NULL);
3567 ovsrec_interface_set_admin_state(if_cfg, NULL);
3568 ovsrec_interface_set_duplex(if_cfg, NULL);
3569 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3570 ovsrec_interface_set_link_state(if_cfg, NULL);
3571 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3572 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3573 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3574 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3575 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3576 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3577 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3581 struct iface_delete_queues_cbdata {
3582 struct netdev *netdev;
3583 const struct ovsdb_datum *queues;
3587 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3589 union ovsdb_atom atom;
3591 atom.integer = target;
3592 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3596 iface_delete_queues(unsigned int queue_id,
3597 const struct smap *details OVS_UNUSED, void *cbdata_)
3599 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3601 if (!queue_ids_include(cbdata->queues, queue_id)) {
3602 netdev_delete_queue(cbdata->netdev, queue_id);
3607 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3609 struct ofpbuf queues_buf;
3611 ofpbuf_init(&queues_buf, 0);
3613 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3614 netdev_set_qos(iface->netdev, NULL, NULL);
3616 struct iface_delete_queues_cbdata cbdata;
3620 /* Configure top-level Qos for 'iface'. */
3621 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3623 /* Deconfigure queues that were deleted. */
3624 cbdata.netdev = iface->netdev;
3625 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3627 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3629 /* Configure queues for 'iface'. */
3631 for (i = 0; i < qos->n_queues; i++) {
3632 const struct ovsrec_queue *queue = qos->value_queues[i];
3633 unsigned int queue_id = qos->key_queues[i];
3635 if (queue_id == 0) {
3639 if (queue->n_dscp == 1) {
3640 struct ofproto_port_queue *port_queue;
3642 port_queue = ofpbuf_put_uninit(&queues_buf,
3643 sizeof *port_queue);
3644 port_queue->queue = queue_id;
3645 port_queue->dscp = queue->dscp[0];
3648 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3651 struct smap details;
3653 smap_init(&details);
3654 netdev_set_queue(iface->netdev, 0, &details);
3655 smap_destroy(&details);
3659 if (iface->ofp_port != OFPP_NONE) {
3660 const struct ofproto_port_queue *port_queues = queues_buf.data;
3661 size_t n_queues = queues_buf.size / sizeof *port_queues;
3663 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3664 port_queues, n_queues);
3667 netdev_set_policing(iface->netdev,
3668 iface->cfg->ingress_policing_rate,
3669 iface->cfg->ingress_policing_burst);
3671 ofpbuf_uninit(&queues_buf);
3675 iface_configure_cfm(struct iface *iface)
3677 const struct ovsrec_interface *cfg = iface->cfg;
3678 const char *opstate_str;
3679 const char *cfm_ccm_vlan;
3680 struct cfm_settings s;
3681 struct smap netdev_args;
3683 if (!cfg->n_cfm_mpid) {
3684 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3688 s.check_tnl_key = false;
3689 smap_init(&netdev_args);
3690 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3691 const char *key = smap_get(&netdev_args, "key");
3692 const char *in_key = smap_get(&netdev_args, "in_key");
3694 s.check_tnl_key = (key && !strcmp(key, "flow"))
3695 || (in_key && !strcmp(in_key, "flow"));
3697 smap_destroy(&netdev_args);
3699 s.mpid = *cfg->cfm_mpid;
3700 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3701 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3702 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3704 if (s.interval <= 0) {
3708 if (!cfm_ccm_vlan) {
3710 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3711 s.ccm_vlan = CFM_RANDOM_VLAN;
3713 s.ccm_vlan = atoi(cfm_ccm_vlan);
3714 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3719 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3721 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3723 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3724 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3726 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3729 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3730 * instead of obtaining it from the database. */
3732 iface_is_synthetic(const struct iface *iface)
3734 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3738 iface_pick_ofport(const struct ovsrec_interface *cfg)
3740 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3742 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3747 /* Port mirroring. */
3749 static struct mirror *
3750 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3754 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3755 if (uuid_equals(uuid, &m->uuid)) {
3763 bridge_configure_mirrors(struct bridge *br)
3765 const struct ovsdb_datum *mc;
3766 unsigned long *flood_vlans;
3767 struct mirror *m, *next;
3770 /* Get rid of deleted mirrors. */
3771 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3772 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3773 union ovsdb_atom atom;
3775 atom.uuid = m->uuid;
3776 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3781 /* Add new mirrors and reconfigure existing ones. */
3782 for (i = 0; i < br->cfg->n_mirrors; i++) {
3783 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3784 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3786 m = mirror_create(br, cfg);
3789 if (!mirror_configure(m)) {
3794 /* Update flooded vlans (for RSPAN). */
3795 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3796 br->cfg->n_flood_vlans);
3797 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3798 bitmap_free(flood_vlans);
3801 static struct mirror *
3802 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3806 m = xzalloc(sizeof *m);
3807 m->uuid = cfg->header_.uuid;
3808 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3810 m->name = xstrdup(cfg->name);
3816 mirror_destroy(struct mirror *m)
3819 struct bridge *br = m->bridge;
3822 ofproto_mirror_unregister(br->ofproto, m);
3825 hmap_remove(&br->mirrors, &m->hmap_node);
3832 mirror_collect_ports(struct mirror *m,
3833 struct ovsrec_port **in_ports, int n_in_ports,
3834 void ***out_portsp, size_t *n_out_portsp)
3836 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3837 size_t n_out_ports = 0;
3840 for (i = 0; i < n_in_ports; i++) {
3841 const char *name = in_ports[i]->name;
3842 struct port *port = port_lookup(m->bridge, name);
3844 out_ports[n_out_ports++] = port;
3846 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3847 "port %s", m->bridge->name, m->name, name);
3850 *out_portsp = out_ports;
3851 *n_out_portsp = n_out_ports;
3855 mirror_configure(struct mirror *m)
3857 const struct ovsrec_mirror *cfg = m->cfg;
3858 struct ofproto_mirror_settings s;
3861 if (strcmp(cfg->name, m->name)) {
3863 m->name = xstrdup(cfg->name);
3867 /* Get output port or VLAN. */
3868 if (cfg->output_port) {
3869 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3870 if (!s.out_bundle) {
3871 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3872 m->bridge->name, m->name);
3875 s.out_vlan = UINT16_MAX;
3877 if (cfg->output_vlan) {
3878 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3879 "output vlan; ignoring output vlan",
3880 m->bridge->name, m->name);
3882 } else if (cfg->output_vlan) {
3883 /* The database should prevent invalid VLAN values. */
3884 s.out_bundle = NULL;
3885 s.out_vlan = *cfg->output_vlan;
3887 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3888 m->bridge->name, m->name);
3892 /* Get port selection. */
3893 if (cfg->select_all) {
3894 size_t n_ports = hmap_count(&m->bridge->ports);
3895 void **ports = xmalloc(n_ports * sizeof *ports);
3900 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3910 /* Get ports, dropping ports that don't exist.
3911 * The IDL ensures that there are no duplicates. */
3912 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3913 &s.srcs, &s.n_srcs);
3914 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3915 &s.dsts, &s.n_dsts);
3918 /* Get VLAN selection. */
3919 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3922 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3925 if (s.srcs != s.dsts) {
3934 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3936 * This is deprecated. It is only for compatibility with broken device drivers
3937 * in old versions of Linux that do not properly support VLANs when VLAN
3938 * devices are not used. When broken device drivers are no longer in
3939 * widespread use, we will delete these interfaces. */
3941 static struct ovsrec_port **recs;
3942 static size_t n_recs, allocated_recs;
3944 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3945 * splinters are reconfigured. */
3947 register_rec(struct ovsrec_port *rec)
3949 if (n_recs >= allocated_recs) {
3950 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3952 recs[n_recs++] = rec;
3955 /* Frees all of the ports registered with register_reg(). */
3957 free_registered_recs(void)
3961 for (i = 0; i < n_recs; i++) {
3962 struct ovsrec_port *port = recs[i];
3965 for (j = 0; j < port->n_interfaces; j++) {
3966 struct ovsrec_interface *iface = port->interfaces[j];
3971 smap_destroy(&port->other_config);
3972 free(port->interfaces);
3980 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3983 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3985 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3989 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3992 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3993 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3994 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3997 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3998 * use, returns NULL.
4000 * Updates 'vlan_splinters_enabled_anywhere'. */
4001 static unsigned long int *
4002 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4004 unsigned long int *splinter_vlans;
4005 struct sset splinter_ifaces;
4006 const char *real_dev_name;
4007 struct shash *real_devs;
4008 struct shash_node *node;
4012 /* Free space allocated for synthesized ports and interfaces, since we're
4013 * in the process of reconstructing all of them. */
4014 free_registered_recs();
4016 splinter_vlans = bitmap_allocate(4096);
4017 sset_init(&splinter_ifaces);
4018 vlan_splinters_enabled_anywhere = false;
4019 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4020 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4023 for (j = 0; j < br_cfg->n_ports; j++) {
4024 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4027 for (k = 0; k < port_cfg->n_interfaces; k++) {
4028 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4030 if (vlan_splinters_is_enabled(iface_cfg)) {
4031 vlan_splinters_enabled_anywhere = true;
4032 sset_add(&splinter_ifaces, iface_cfg->name);
4033 vlan_bitmap_from_array__(port_cfg->trunks,
4039 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4040 bitmap_set1(splinter_vlans, *port_cfg->tag);
4045 if (!vlan_splinters_enabled_anywhere) {
4046 free(splinter_vlans);
4047 sset_destroy(&splinter_ifaces);
4051 HMAP_FOR_EACH (br, node, &all_bridges) {
4053 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4057 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4058 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4059 * device to be created for it. */
4060 bitmap_set0(splinter_vlans, 0);
4061 bitmap_set0(splinter_vlans, 4095);
4063 /* Delete all VLAN devices that we don't need. */
4065 real_devs = vlandev_get_real_devs();
4066 SHASH_FOR_EACH (node, real_devs) {
4067 const struct vlan_real_dev *real_dev = node->data;
4068 const struct vlan_dev *vlan_dev;
4069 bool real_dev_has_splinters;
4071 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4073 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4074 if (!real_dev_has_splinters
4075 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4076 struct netdev *netdev;
4078 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4079 if (!netdev_get_in4(netdev, NULL, NULL) ||
4080 !netdev_get_in6(netdev, NULL)) {
4081 vlandev_del(vlan_dev->name);
4083 /* It has an IP address configured, so we don't own
4084 * it. Don't delete it. */
4086 netdev_close(netdev);
4093 /* Add all VLAN devices that we need. */
4094 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4097 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4098 if (!vlandev_get_name(real_dev_name, vid)) {
4099 vlandev_add(real_dev_name, vid);
4106 sset_destroy(&splinter_ifaces);
4108 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4109 free(splinter_vlans);
4112 return splinter_vlans;
4115 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4118 configure_splinter_port(struct port *port)
4120 struct ofproto *ofproto = port->bridge->ofproto;
4121 ofp_port_t realdev_ofp_port;
4122 const char *realdev_name;
4123 struct iface *vlandev, *realdev;
4125 ofproto_bundle_unregister(port->bridge->ofproto, port);
4127 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4130 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4131 realdev = iface_lookup(port->bridge, realdev_name);
4132 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4134 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4138 static struct ovsrec_port *
4139 synthesize_splinter_port(const char *real_dev_name,
4140 const char *vlan_dev_name, int vid)
4142 struct ovsrec_interface *iface;
4143 struct ovsrec_port *port;
4145 iface = xmalloc(sizeof *iface);
4146 ovsrec_interface_init(iface);
4147 iface->name = xstrdup(vlan_dev_name);
4148 iface->type = "system";
4150 port = xmalloc(sizeof *port);
4151 ovsrec_port_init(port);
4152 port->interfaces = xmemdup(&iface, sizeof iface);
4153 port->n_interfaces = 1;
4154 port->name = xstrdup(vlan_dev_name);
4155 port->vlan_mode = "splinter";
4156 port->tag = xmalloc(sizeof *port->tag);
4159 smap_add(&port->other_config, "realdev", real_dev_name);
4165 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4166 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4167 * 1-bit in the 'splinter_vlans' bitmap. */
4169 add_vlan_splinter_ports(struct bridge *br,
4170 const unsigned long int *splinter_vlans,
4171 struct shash *ports)
4175 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4176 * we're modifying 'ports'. */
4177 for (i = 0; i < br->cfg->n_ports; i++) {
4178 const char *name = br->cfg->ports[i]->name;
4179 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4182 for (j = 0; j < port_cfg->n_interfaces; j++) {
4183 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4185 if (vlan_splinters_is_enabled(iface_cfg)) {
4186 const char *real_dev_name;
4189 real_dev_name = iface_cfg->name;
4190 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4191 const char *vlan_dev_name;
4193 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4195 && !shash_find(ports, vlan_dev_name)) {
4196 shash_add(ports, vlan_dev_name,
4197 synthesize_splinter_port(
4198 real_dev_name, vlan_dev_name, vid));
4207 mirror_refresh_stats(struct mirror *m)
4209 struct ofproto *ofproto = m->bridge->ofproto;
4210 uint64_t tx_packets, tx_bytes;
4213 size_t stat_cnt = 0;
4215 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4216 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4220 if (tx_packets != UINT64_MAX) {
4221 keys[stat_cnt] = "tx_packets";
4222 values[stat_cnt] = tx_packets;
4225 if (tx_bytes != UINT64_MAX) {
4226 keys[stat_cnt] = "tx_bytes";
4227 values[stat_cnt] = tx_bytes;
4231 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);