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,
698 ovs_strerror(error));
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,
707 ovs_strerror(error));
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 } else if (strcmp(opt, "without-facets")) {
820 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
824 ofproto_set_flow_miss_model(model);
827 /* Pick local port hardware address and datapath ID for 'br'. */
829 bridge_configure_datapath_id(struct bridge *br)
831 uint8_t ea[ETH_ADDR_LEN];
833 struct iface *local_iface;
834 struct iface *hw_addr_iface;
837 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
838 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
840 int error = netdev_set_etheraddr(local_iface->netdev, ea);
842 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
843 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
844 "Ethernet address: %s",
845 br->name, ovs_strerror(error));
848 memcpy(br->ea, ea, ETH_ADDR_LEN);
850 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
851 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
852 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
853 ofproto_set_datapath_id(br->ofproto, dpid);
856 dpid_string = xasprintf("%016"PRIx64, dpid);
857 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
861 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
864 bridge_get_allowed_versions(struct bridge *br)
866 if (!br->cfg->n_protocols)
869 return ofputil_versions_from_strings(br->cfg->protocols,
870 br->cfg->n_protocols);
873 /* Set NetFlow configuration on 'br'. */
875 bridge_configure_netflow(struct bridge *br)
877 struct ovsrec_netflow *cfg = br->cfg->netflow;
878 struct netflow_options opts;
881 ofproto_set_netflow(br->ofproto, NULL);
885 memset(&opts, 0, sizeof opts);
887 /* Get default NetFlow configuration from datapath.
888 * Apply overrides from 'cfg'. */
889 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
890 if (cfg->engine_type) {
891 opts.engine_type = *cfg->engine_type;
893 if (cfg->engine_id) {
894 opts.engine_id = *cfg->engine_id;
897 /* Configure active timeout interval. */
898 opts.active_timeout = cfg->active_timeout;
899 if (!opts.active_timeout) {
900 opts.active_timeout = -1;
901 } else if (opts.active_timeout < 0) {
902 VLOG_WARN("bridge %s: active timeout interval set to negative "
903 "value, using default instead (%d seconds)", br->name,
904 NF_ACTIVE_TIMEOUT_DEFAULT);
905 opts.active_timeout = -1;
908 /* Add engine ID to interface number to disambiguate bridgs? */
909 opts.add_id_to_iface = cfg->add_id_to_interface;
910 if (opts.add_id_to_iface) {
911 if (opts.engine_id > 0x7f) {
912 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
913 "another vswitch, choose an engine id less than 128",
916 if (hmap_count(&br->ports) > 508) {
917 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
918 "another port when more than 508 ports are used",
924 sset_init(&opts.collectors);
925 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
928 if (ofproto_set_netflow(br->ofproto, &opts)) {
929 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
931 sset_destroy(&opts.collectors);
934 /* Set sFlow configuration on 'br'. */
936 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
938 const struct ovsrec_sflow *cfg = br->cfg->sflow;
939 struct ovsrec_controller **controllers;
940 struct ofproto_sflow_options oso;
941 size_t n_controllers;
945 ofproto_set_sflow(br->ofproto, NULL);
949 memset(&oso, 0, sizeof oso);
951 sset_init(&oso.targets);
952 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
954 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
956 oso.sampling_rate = *cfg->sampling;
959 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
961 oso.polling_interval = *cfg->polling;
964 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
966 oso.header_len = *cfg->header;
969 oso.sub_id = (*sflow_bridge_number)++;
970 oso.agent_device = cfg->agent;
972 oso.control_ip = NULL;
973 n_controllers = bridge_get_controllers(br, &controllers);
974 for (i = 0; i < n_controllers; i++) {
975 if (controllers[i]->local_ip) {
976 oso.control_ip = controllers[i]->local_ip;
980 ofproto_set_sflow(br->ofproto, &oso);
982 sset_destroy(&oso.targets);
985 /* Set IPFIX configuration on 'br'. */
987 bridge_configure_ipfix(struct bridge *br)
989 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
990 const struct ovsrec_flow_sample_collector_set *fe_cfg;
991 struct ofproto_ipfix_bridge_exporter_options be_opts;
992 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
993 size_t n_fe_opts = 0;
995 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
996 if (fe_cfg->bridge == br->cfg) {
1001 if (!be_cfg && n_fe_opts == 0) {
1002 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1007 memset(&be_opts, 0, sizeof be_opts);
1009 sset_init(&be_opts.targets);
1010 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1012 if (be_cfg->sampling) {
1013 be_opts.sampling_rate = *be_cfg->sampling;
1015 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1017 if (be_cfg->obs_domain_id) {
1018 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1020 if (be_cfg->obs_point_id) {
1021 be_opts.obs_point_id = *be_cfg->obs_point_id;
1025 if (n_fe_opts > 0) {
1026 struct ofproto_ipfix_flow_exporter_options *opts;
1027 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1029 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1030 if (fe_cfg->bridge == br->cfg) {
1031 opts->collector_set_id = fe_cfg->id;
1032 sset_init(&opts->targets);
1033 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1034 fe_cfg->ipfix->n_targets);
1040 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1044 sset_destroy(&be_opts.targets);
1047 if (n_fe_opts > 0) {
1048 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1050 for (i = 0; i < n_fe_opts; i++) {
1051 sset_destroy(&opts->targets);
1059 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1060 struct ofproto_port_stp_settings *port_s,
1061 int *port_num_counter, unsigned long *port_num_bitmap)
1063 const char *config_str;
1064 struct iface *iface;
1066 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1067 port_s->enable = false;
1070 port_s->enable = true;
1073 /* STP over bonds is not supported. */
1074 if (!list_is_singleton(&port->ifaces)) {
1075 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1077 port_s->enable = false;
1081 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1083 /* Internal ports shouldn't participate in spanning tree, so
1085 if (!strcmp(iface->type, "internal")) {
1086 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1087 port_s->enable = false;
1091 /* STP on mirror output ports is not supported. */
1092 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1093 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1094 port_s->enable = false;
1098 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1100 unsigned long int port_num = strtoul(config_str, NULL, 0);
1101 int port_idx = port_num - 1;
1103 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1104 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1105 port_s->enable = false;
1109 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1110 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1111 port->name, port_num);
1112 port_s->enable = false;
1115 bitmap_set1(port_num_bitmap, port_idx);
1116 port_s->port_num = port_idx;
1118 if (*port_num_counter >= STP_MAX_PORTS) {
1119 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1120 port_s->enable = false;
1124 port_s->port_num = (*port_num_counter)++;
1127 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1129 port_s->path_cost = strtoul(config_str, NULL, 10);
1131 enum netdev_features current;
1134 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1135 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1136 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1139 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1141 port_s->priority = strtoul(config_str, NULL, 0);
1143 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1147 /* Set spanning tree configuration on 'br'. */
1149 bridge_configure_stp(struct bridge *br)
1151 if (!br->cfg->stp_enable) {
1152 ofproto_set_stp(br->ofproto, NULL);
1154 struct ofproto_stp_settings br_s;
1155 const char *config_str;
1157 int port_num_counter;
1158 unsigned long *port_num_bitmap;
1160 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1162 uint8_t ea[ETH_ADDR_LEN];
1164 if (eth_addr_from_string(config_str, ea)) {
1165 br_s.system_id = eth_addr_to_uint64(ea);
1167 br_s.system_id = eth_addr_to_uint64(br->ea);
1168 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1169 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1172 br_s.system_id = eth_addr_to_uint64(br->ea);
1175 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1177 br_s.priority = strtoul(config_str, NULL, 0);
1179 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1182 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1184 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1186 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1189 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1191 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1193 br_s.max_age = STP_DEFAULT_MAX_AGE;
1196 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1198 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1200 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1203 /* Configure STP on the bridge. */
1204 if (ofproto_set_stp(br->ofproto, &br_s)) {
1205 VLOG_ERR("bridge %s: could not enable STP", br->name);
1209 /* Users must either set the port number with the "stp-port-num"
1210 * configuration on all ports or none. If manual configuration
1211 * is not done, then we allocate them sequentially. */
1212 port_num_counter = 0;
1213 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1214 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1215 struct ofproto_port_stp_settings port_s;
1216 struct iface *iface;
1218 port_configure_stp(br->ofproto, port, &port_s,
1219 &port_num_counter, port_num_bitmap);
1221 /* As bonds are not supported, just apply configuration to
1222 * all interfaces. */
1223 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1224 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1226 VLOG_ERR("port %s: could not enable STP", port->name);
1232 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1233 && port_num_counter) {
1234 VLOG_ERR("bridge %s: must manually configure all STP port "
1235 "IDs or none, disabling", br->name);
1236 ofproto_set_stp(br->ofproto, NULL);
1238 bitmap_free(port_num_bitmap);
1243 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1245 const struct port *port = port_lookup(br, name);
1246 return port && port_is_bond_fake_iface(port);
1250 port_is_bond_fake_iface(const struct port *port)
1252 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1256 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1258 struct bridge *br, *next;
1259 struct shash new_br;
1262 /* Collect new bridges' names and types. */
1263 shash_init(&new_br);
1264 for (i = 0; i < cfg->n_bridges; i++) {
1265 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1266 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1268 if (strchr(br_cfg->name, '/')) {
1269 /* Prevent remote ovsdb-server users from accessing arbitrary
1270 * directories, e.g. consider a bridge named "../../../etc/". */
1271 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1273 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1274 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1278 /* Get rid of deleted bridges or those whose types have changed.
1279 * Update 'cfg' of bridges that still exist. */
1280 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1281 br->cfg = shash_find_data(&new_br, br->name);
1282 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1283 br->cfg->datapath_type))) {
1288 /* Add new bridges. */
1289 for (i = 0; i < cfg->n_bridges; i++) {
1290 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1291 struct bridge *br = bridge_lookup(br_cfg->name);
1293 bridge_create(br_cfg);
1297 shash_destroy(&new_br);
1301 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1303 struct bridge *br = iface->port->bridge;
1305 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1306 iface->ofp_port = ofp_port;
1307 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1308 hash_ofp_port(ofp_port));
1309 iface_set_ofport(iface->cfg, ofp_port);
1312 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1313 * Returns 0 if successful, otherwise a positive errno value. */
1315 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1316 struct netdev *netdev)
1320 error = netdev_set_config(netdev, &iface_cfg->options);
1322 VLOG_WARN("could not configure network device %s (%s)",
1323 iface_cfg->name, ovs_strerror(error));
1328 /* This function determines whether 'ofproto_port', which is attached to
1329 * br->ofproto's datapath, is one that we want in 'br'.
1331 * If it is, it returns true, after creating an iface (if necessary),
1332 * configuring the iface's netdev according to the iface's options, and setting
1333 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1335 * If, on the other hand, 'port' should be removed, it returns false. The
1336 * caller should later detach the port from br->ofproto. */
1338 bridge_refresh_one_ofp_port(struct bridge *br,
1339 const struct ofproto_port *ofproto_port)
1341 const char *name = ofproto_port->name;
1342 const char *type = ofproto_port->type;
1343 ofp_port_t ofp_port = ofproto_port->ofp_port;
1345 struct iface *iface = iface_lookup(br, name);
1347 /* Check that the name-to-number mapping is one-to-one. */
1348 if (iface->ofp_port != OFPP_NONE) {
1349 VLOG_WARN("bridge %s: interface %s reported twice",
1352 } else if (iface_from_ofp_port(br, ofp_port)) {
1353 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1354 br->name, ofp_port);
1358 /* There's a configured interface named 'name'. */
1359 if (strcmp(type, iface->type)
1360 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1361 /* It's the wrong type, or it's the right type but can't be
1362 * configured as the user requested, so we must destroy it. */
1365 /* It's the right type and configured correctly. Keep it. */
1366 iface_set_ofp_port(iface, ofp_port);
1369 } else if (bridge_has_bond_fake_iface(br, name)
1370 && !strcmp(type, "internal")) {
1371 /* It's a bond fake iface. Keep it. */
1374 /* There's no configured interface named 'name', but there might be an
1375 * interface of that name queued to be created.
1377 * If there is, and it has the correct type, then try to configure it
1378 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1379 * delete it and later try to re-add it. */
1380 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1382 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1383 && iface_create(br, if_cfg, ofp_port));
1387 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1388 * consistent with the ofp_ports in "br->ofproto". */
1390 bridge_refresh_ofp_port(struct bridge *br)
1392 struct ofproto_port_dump dump;
1393 struct ofproto_port ofproto_port;
1394 struct port *port, *port_next;
1396 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1397 hmap_clear(&br->ifaces);
1398 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1399 struct iface *iface;
1401 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1402 iface->ofp_port = OFPP_NONE;
1406 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1407 * already exist in the datapath and promote them to full fledged "struct
1408 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1409 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1410 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1411 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1412 garbage->ofp_port = ofproto_port.ofp_port;
1413 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1417 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1418 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1419 * later they can be added to ofproto. */
1420 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1421 struct iface *iface, *iface_next;
1423 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1424 if (iface->ofp_port == OFPP_NONE) {
1425 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1426 iface_destroy(iface);
1430 if (list_is_empty(&port->ifaces)) {
1436 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1437 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1438 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1441 * If successful, returns 0 and stores the network device in '*netdevp'. On
1442 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1444 iface_do_create(const struct bridge *br,
1445 const struct if_cfg *if_cfg,
1446 ofp_port_t *ofp_portp, struct netdev **netdevp)
1448 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1449 const struct ovsrec_port *port_cfg = if_cfg->parent;
1450 struct netdev *netdev = NULL;
1453 if (netdev_is_reserved_name(iface_cfg->name)) {
1454 VLOG_WARN("could not create interface %s, name is reserved",
1460 error = netdev_open(iface_cfg->name,
1461 iface_get_type(iface_cfg, br->cfg), &netdev);
1463 VLOG_WARN("could not open network device %s (%s)",
1464 iface_cfg->name, ovs_strerror(error));
1468 error = iface_set_netdev_config(iface_cfg, netdev);
1473 if (*ofp_portp == OFPP_NONE) {
1474 ofp_port_t ofp_port = if_cfg->ofport;
1476 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1480 *ofp_portp = ofp_port;
1482 VLOG_INFO("bridge %s: added interface %s on port %d",
1483 br->name, iface_cfg->name, *ofp_portp);
1485 VLOG_DBG("bridge %s: interface %s is on port %d",
1486 br->name, iface_cfg->name, *ofp_portp);
1489 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1490 || iface_is_internal(iface_cfg, br->cfg)) {
1491 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1499 netdev_close(netdev);
1503 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1504 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1505 * automatically allocated for the iface. Takes ownership of and
1506 * deallocates 'if_cfg'.
1508 * Return true if an iface is successfully created, false otherwise. */
1510 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1512 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1513 const struct ovsrec_port *port_cfg = if_cfg->parent;
1515 struct netdev *netdev;
1516 struct iface *iface;
1521 /* Do the bits that can fail up front.
1523 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1524 * internal datastructures may not be consistent. Eventually, when port
1525 * additions and deletions are cheaper, these calls should be removed. */
1527 ovs_assert(!iface_lookup(br, iface_cfg->name));
1528 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1531 iface_set_ofport(iface_cfg, OFPP_NONE);
1532 iface_clear_db_record(iface_cfg);
1537 /* Get or create the port structure. */
1538 port = port_lookup(br, port_cfg->name);
1540 port = port_create(br, port_cfg);
1543 /* Create the iface structure. */
1544 iface = xzalloc(sizeof *iface);
1545 list_push_back(&port->ifaces, &iface->port_elem);
1546 hmap_insert(&br->iface_by_name, &iface->name_node,
1547 hash_string(iface_cfg->name, 0));
1549 iface->name = xstrdup(iface_cfg->name);
1550 iface->ofp_port = OFPP_NONE;
1551 iface->netdev = netdev;
1552 iface->type = iface_get_type(iface_cfg, br->cfg);
1553 iface->cfg = iface_cfg;
1555 iface_set_ofp_port(iface, ofp_port);
1557 /* Populate initial status in database. */
1558 iface_refresh_stats(iface);
1559 iface_refresh_status(iface);
1561 /* Add bond fake iface if necessary. */
1562 if (port_is_bond_fake_iface(port)) {
1563 struct ofproto_port ofproto_port;
1565 if (ofproto_port_query_by_name(br->ofproto, port->name,
1567 struct netdev *netdev;
1570 error = netdev_open(port->name, "internal", &netdev);
1572 ofp_port_t fake_ofp_port = if_cfg->ofport;
1574 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1575 netdev_close(netdev);
1577 VLOG_WARN("could not open network device %s (%s)",
1578 port->name, ovs_strerror(error));
1581 /* Already exists, nothing to do. */
1582 ofproto_port_destroy(&ofproto_port);
1587 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1593 /* Set forward BPDU option. */
1595 bridge_configure_forward_bpdu(struct bridge *br)
1597 ofproto_set_forward_bpdu(br->ofproto,
1598 smap_get_bool(&br->cfg->other_config,
1603 /* Set MAC learning table configuration for 'br'. */
1605 bridge_configure_mac_table(struct bridge *br)
1607 const char *idle_time_str;
1610 const char *mac_table_size_str;
1613 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1614 idle_time = (idle_time_str && atoi(idle_time_str)
1615 ? atoi(idle_time_str)
1616 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1618 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1619 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1620 ? atoi(mac_table_size_str)
1623 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1627 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1628 struct iface **hw_addr_iface)
1630 struct hmapx mirror_output_ports;
1633 bool found_addr = false;
1637 *hw_addr_iface = NULL;
1639 /* Did the user request a particular MAC? */
1640 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1641 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1642 if (eth_addr_is_multicast(ea)) {
1643 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1644 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1645 } else if (eth_addr_is_zero(ea)) {
1646 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1652 /* Mirror output ports don't participate in picking the local hardware
1653 * address. ofproto can't help us find out whether a given port is a
1654 * mirror output because we haven't configured mirrors yet, so we need to
1655 * accumulate them ourselves. */
1656 hmapx_init(&mirror_output_ports);
1657 for (i = 0; i < br->cfg->n_mirrors; i++) {
1658 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1659 if (m->output_port) {
1660 hmapx_add(&mirror_output_ports, m->output_port);
1664 /* Otherwise choose the minimum non-local MAC address among all of the
1666 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1667 uint8_t iface_ea[ETH_ADDR_LEN];
1668 struct iface *candidate;
1669 struct iface *iface;
1671 /* Mirror output ports don't participate. */
1672 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1676 /* Choose the MAC address to represent the port. */
1678 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1679 /* Find the interface with this Ethernet address (if any) so that
1680 * we can provide the correct devname to the caller. */
1681 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1682 uint8_t candidate_ea[ETH_ADDR_LEN];
1683 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1684 && eth_addr_equals(iface_ea, candidate_ea)) {
1689 /* Choose the interface whose MAC address will represent the port.
1690 * The Linux kernel bonding code always chooses the MAC address of
1691 * the first slave added to a bond, and the Fedora networking
1692 * scripts always add slaves to a bond in alphabetical order, so
1693 * for compatibility we choose the interface with the name that is
1694 * first in alphabetical order. */
1695 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1696 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1701 /* The local port doesn't count (since we're trying to choose its
1702 * MAC address anyway). */
1703 if (iface->ofp_port == OFPP_LOCAL) {
1708 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1714 /* Compare against our current choice. */
1715 if (!eth_addr_is_multicast(iface_ea) &&
1716 !eth_addr_is_local(iface_ea) &&
1717 !eth_addr_is_reserved(iface_ea) &&
1718 !eth_addr_is_zero(iface_ea) &&
1719 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1721 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1722 *hw_addr_iface = iface;
1728 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1729 *hw_addr_iface = NULL;
1732 hmapx_destroy(&mirror_output_ports);
1735 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1736 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1737 * an interface on 'br', then that interface must be passed in as
1738 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1739 * 'hw_addr_iface' must be passed in as a null pointer. */
1741 bridge_pick_datapath_id(struct bridge *br,
1742 const uint8_t bridge_ea[ETH_ADDR_LEN],
1743 struct iface *hw_addr_iface)
1746 * The procedure for choosing a bridge MAC address will, in the most
1747 * ordinary case, also choose a unique MAC that we can use as a datapath
1748 * ID. In some special cases, though, multiple bridges will end up with
1749 * the same MAC address. This is OK for the bridges, but it will confuse
1750 * the OpenFlow controller, because each datapath needs a unique datapath
1753 * Datapath IDs must be unique. It is also very desirable that they be
1754 * stable from one run to the next, so that policy set on a datapath
1757 const char *datapath_id;
1760 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1761 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1765 if (!hw_addr_iface) {
1767 * A purely internal bridge, that is, one that has no non-virtual
1768 * network devices on it at all, is difficult because it has no
1769 * natural unique identifier at all.
1771 * When the host is a XenServer, we handle this case by hashing the
1772 * host's UUID with the name of the bridge. Names of bridges are
1773 * persistent across XenServer reboots, although they can be reused if
1774 * an internal network is destroyed and then a new one is later
1775 * created, so this is fairly effective.
1777 * When the host is not a XenServer, we punt by using a random MAC
1778 * address on each run.
1780 const char *host_uuid = xenserver_get_host_uuid();
1782 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1783 dpid = dpid_from_hash(combined, strlen(combined));
1789 return eth_addr_to_uint64(bridge_ea);
1793 dpid_from_hash(const void *data, size_t n)
1795 uint8_t hash[SHA1_DIGEST_SIZE];
1797 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1798 sha1_bytes(data, n, hash);
1799 eth_addr_mark_random(hash);
1800 return eth_addr_to_uint64(hash);
1804 iface_refresh_status(struct iface *iface)
1808 enum netdev_features current;
1812 uint8_t mac[ETH_ADDR_LEN];
1815 if (iface_is_synthetic(iface)) {
1821 if (!netdev_get_status(iface->netdev, &smap)) {
1822 ovsrec_interface_set_status(iface->cfg, &smap);
1824 ovsrec_interface_set_status(iface->cfg, NULL);
1827 smap_destroy(&smap);
1829 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1830 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1832 ovsrec_interface_set_duplex(iface->cfg,
1833 netdev_features_is_full_duplex(current)
1835 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1837 ovsrec_interface_set_duplex(iface->cfg, NULL);
1838 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1841 error = netdev_get_mtu(iface->netdev, &mtu);
1844 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1846 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1849 error = netdev_get_etheraddr(iface->netdev, mac);
1851 char mac_string[32];
1853 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1854 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1856 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1860 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1863 iface_refresh_cfm_stats(struct iface *iface)
1865 const struct ovsrec_interface *cfg = iface->cfg;
1866 struct ofproto_cfm_status status;
1868 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1869 iface->ofp_port, &status)) {
1870 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1871 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1872 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1873 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1874 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1876 const char *reasons[CFM_FAULT_N_REASONS];
1877 int64_t cfm_health = status.health;
1878 bool faulted = status.faults != 0;
1881 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1884 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1885 int reason = 1 << i;
1886 if (status.faults & reason) {
1887 reasons[j++] = cfm_fault_reason_to_str(reason);
1890 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1892 if (status.remote_opstate >= 0) {
1893 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1894 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1896 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1899 ovsrec_interface_set_cfm_remote_mpids(cfg,
1900 (const int64_t *)status.rmps,
1902 if (cfm_health >= 0) {
1903 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1905 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1911 iface_refresh_stats(struct iface *iface)
1913 #define IFACE_STATS \
1914 IFACE_STAT(rx_packets, "rx_packets") \
1915 IFACE_STAT(tx_packets, "tx_packets") \
1916 IFACE_STAT(rx_bytes, "rx_bytes") \
1917 IFACE_STAT(tx_bytes, "tx_bytes") \
1918 IFACE_STAT(rx_dropped, "rx_dropped") \
1919 IFACE_STAT(tx_dropped, "tx_dropped") \
1920 IFACE_STAT(rx_errors, "rx_errors") \
1921 IFACE_STAT(tx_errors, "tx_errors") \
1922 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1923 IFACE_STAT(rx_over_errors, "rx_over_err") \
1924 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1925 IFACE_STAT(collisions, "collisions")
1927 #define IFACE_STAT(MEMBER, NAME) + 1
1928 enum { N_IFACE_STATS = IFACE_STATS };
1930 int64_t values[N_IFACE_STATS];
1931 char *keys[N_IFACE_STATS];
1934 struct netdev_stats stats;
1936 if (iface_is_synthetic(iface)) {
1940 /* Intentionally ignore return value, since errors will set 'stats' to
1941 * all-1s, and we will deal with that correctly below. */
1942 netdev_get_stats(iface->netdev, &stats);
1944 /* Copy statistics into keys[] and values[]. */
1946 #define IFACE_STAT(MEMBER, NAME) \
1947 if (stats.MEMBER != UINT64_MAX) { \
1949 values[n] = stats.MEMBER; \
1954 ovs_assert(n <= N_IFACE_STATS);
1956 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1961 br_refresh_stp_status(struct bridge *br)
1963 struct smap smap = SMAP_INITIALIZER(&smap);
1964 struct ofproto *ofproto = br->ofproto;
1965 struct ofproto_stp_status status;
1967 if (ofproto_get_stp_status(ofproto, &status)) {
1971 if (!status.enabled) {
1972 ovsrec_bridge_set_status(br->cfg, NULL);
1976 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1977 STP_ID_ARGS(status.bridge_id));
1978 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1979 STP_ID_ARGS(status.designated_root));
1980 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1982 ovsrec_bridge_set_status(br->cfg, &smap);
1983 smap_destroy(&smap);
1987 port_refresh_stp_status(struct port *port)
1989 struct ofproto *ofproto = port->bridge->ofproto;
1990 struct iface *iface;
1991 struct ofproto_port_stp_status status;
1993 int64_t int_values[3];
1996 if (port_is_synthetic(port)) {
2000 /* STP doesn't currently support bonds. */
2001 if (!list_is_singleton(&port->ifaces)) {
2002 ovsrec_port_set_status(port->cfg, NULL);
2006 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2008 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2012 if (!status.enabled) {
2013 ovsrec_port_set_status(port->cfg, NULL);
2014 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2018 /* Set Status column. */
2020 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2021 smap_add(&smap, "stp_state", stp_state_name(status.state));
2022 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2023 smap_add(&smap, "stp_role", stp_role_name(status.role));
2024 ovsrec_port_set_status(port->cfg, &smap);
2025 smap_destroy(&smap);
2027 /* Set Statistics column. */
2028 keys[0] = "stp_tx_count";
2029 int_values[0] = status.tx_count;
2030 keys[1] = "stp_rx_count";
2031 int_values[1] = status.rx_count;
2032 keys[2] = "stp_error_count";
2033 int_values[2] = status.error_count;
2035 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2036 ARRAY_SIZE(int_values));
2040 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2042 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2046 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2048 bool enable = enable_system_stats(cfg);
2050 system_stats_enable(enable);
2052 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2057 run_system_stats(void)
2059 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2062 stats = system_stats_run();
2064 struct ovsdb_idl_txn *txn;
2065 struct ovsdb_datum datum;
2067 txn = ovsdb_idl_txn_create(idl);
2068 ovsdb_datum_from_smap(&datum, stats);
2069 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2071 ovsdb_idl_txn_commit(txn);
2072 ovsdb_idl_txn_destroy(txn);
2078 static inline const char *
2079 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2082 case OFPCR12_ROLE_EQUAL:
2084 case OFPCR12_ROLE_MASTER:
2086 case OFPCR12_ROLE_SLAVE:
2088 case OFPCR12_ROLE_NOCHANGE:
2090 return "*** INVALID ROLE ***";
2095 refresh_controller_status(void)
2099 const struct ovsrec_controller *cfg;
2103 /* Accumulate status for controllers on all bridges. */
2104 HMAP_FOR_EACH (br, node, &all_bridges) {
2105 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2108 /* Update each controller in the database with current status. */
2109 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2110 struct ofproto_controller_info *cinfo =
2111 shash_find_data(&info, cfg->target);
2114 struct smap smap = SMAP_INITIALIZER(&smap);
2115 const char **values = cinfo->pairs.values;
2116 const char **keys = cinfo->pairs.keys;
2119 for (i = 0; i < cinfo->pairs.n; i++) {
2120 smap_add(&smap, keys[i], values[i]);
2123 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2124 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2126 ovsrec_controller_set_status(cfg, &smap);
2127 smap_destroy(&smap);
2129 ovsrec_controller_set_is_connected(cfg, false);
2130 ovsrec_controller_set_role(cfg, NULL);
2131 ovsrec_controller_set_status(cfg, NULL);
2135 ofproto_free_ofproto_controller_info(&info);
2140 * Some information in the database must be kept as up-to-date as possible to
2141 * allow controllers to respond rapidly to network outages. We call these
2142 * statistics "instant" stats.
2144 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2145 * assuming that they've changed. The only means we have to determine whether
2146 * they have changed are:
2148 * - Try to commit changes to the database. If nothing changed, then
2149 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2152 * - instant_stats_run() is called late in the run loop, after anything that
2153 * might change any of the instant stats.
2155 * We use these two facts together to avoid waking the process up every
2156 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2159 /* Minimum interval between writing updates to the instant stats to the
2161 #define INSTANT_INTERVAL_MSEC 100
2163 /* Current instant stats database transaction, NULL if there is no ongoing
2165 static struct ovsdb_idl_txn *instant_txn;
2167 /* Next time (in msec on monotonic clock) at which we will update the instant
2169 static long long int instant_next_txn = LLONG_MIN;
2171 /* True if the run loop has run since we last saw that the instant stats were
2172 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2173 * to refresh the instant stats. */
2174 static bool instant_stats_could_have_changed;
2177 instant_stats_run(void)
2179 enum ovsdb_idl_txn_status status;
2181 instant_stats_could_have_changed = true;
2186 if (time_msec() < instant_next_txn) {
2189 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2191 instant_txn = ovsdb_idl_txn_create(idl);
2192 HMAP_FOR_EACH (br, node, &all_bridges) {
2193 struct iface *iface;
2196 br_refresh_stp_status(br);
2198 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2199 port_refresh_stp_status(port);
2202 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2203 enum netdev_flags flags;
2205 const char *link_state;
2206 int64_t link_resets;
2209 if (iface_is_synthetic(iface)) {
2213 current = ofproto_port_is_lacp_current(br->ofproto,
2217 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2219 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2222 error = netdev_get_flags(iface->netdev, &flags);
2224 const char *state = flags & NETDEV_UP ? "up" : "down";
2225 ovsrec_interface_set_admin_state(iface->cfg, state);
2227 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2230 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2231 ovsrec_interface_set_link_state(iface->cfg, link_state);
2233 link_resets = netdev_get_carrier_resets(iface->netdev);
2234 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2236 iface_refresh_cfm_stats(iface);
2239 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2241 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2242 smap_destroy(&smap);
2248 status = ovsdb_idl_txn_commit(instant_txn);
2249 if (status != TXN_INCOMPLETE) {
2250 ovsdb_idl_txn_destroy(instant_txn);
2253 if (status == TXN_UNCHANGED) {
2254 instant_stats_could_have_changed = false;
2259 instant_stats_wait(void)
2262 ovsdb_idl_txn_wait(instant_txn);
2263 } else if (instant_stats_could_have_changed) {
2264 poll_timer_wait_until(instant_next_txn);
2268 /* Performs periodic activity required by bridges that needs to be done with
2269 * the least possible latency.
2271 * It makes sense to call this function a couple of times per poll loop, to
2272 * provide a significant performance boost on some benchmarks with ofprotos
2273 * that use the ofproto-dpif implementation. */
2275 bridge_run_fast(void)
2282 ofproto_enumerate_types(&types);
2283 SSET_FOR_EACH (type, &types) {
2284 ofproto_type_run_fast(type);
2286 sset_destroy(&types);
2288 HMAP_FOR_EACH (br, node, &all_bridges) {
2289 ofproto_run_fast(br->ofproto);
2296 static struct ovsrec_open_vswitch null_cfg;
2297 const struct ovsrec_open_vswitch *cfg;
2298 struct ovsdb_idl_txn *reconf_txn = NULL;
2302 bool vlan_splinters_changed;
2305 ovsrec_open_vswitch_init(&null_cfg);
2307 /* (Re)configure if necessary. */
2308 if (!reconfiguring) {
2311 if (ovsdb_idl_is_lock_contended(idl)) {
2312 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2313 struct bridge *br, *next_br;
2315 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2316 "disabling this process (pid %ld) until it goes away",
2317 (long int) getpid());
2319 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2322 /* Since we will not be running system_stats_run() in this process
2323 * with the current situation of multiple ovs-vswitchd daemons,
2324 * disable system stats collection. */
2325 system_stats_enable(false);
2327 } else if (!ovsdb_idl_has_lock(idl)) {
2331 cfg = ovsrec_open_vswitch_first(idl);
2333 /* Initialize the ofproto library. This only needs to run once, but
2334 * it must be done after the configuration is set. If the
2335 * initialization has already occurred, bridge_init_ofproto()
2336 * returns immediately. */
2337 bridge_init_ofproto(cfg);
2339 /* Once the value of flow-restore-wait is false, we no longer should
2340 * check its value from the database. */
2341 if (cfg && ofproto_get_flow_restore_wait()) {
2342 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2343 "flow-restore-wait", false));
2346 /* Let each datapath type do the work that it needs to do. */
2348 ofproto_enumerate_types(&types);
2349 SSET_FOR_EACH (type, &types) {
2350 ofproto_type_run(type);
2352 sset_destroy(&types);
2354 /* Let each bridge do the work that it needs to do. */
2355 HMAP_FOR_EACH (br, node, &all_bridges) {
2356 ofproto_run(br->ofproto);
2359 /* Re-configure SSL. We do this on every trip through the main loop,
2360 * instead of just when the database changes, because the contents of the
2361 * key and certificate files can change without the database changing.
2363 * We do this before bridge_reconfigure() because that function might
2364 * initiate SSL connections and thus requires SSL to be configured. */
2365 if (cfg && cfg->ssl) {
2366 const struct ovsrec_ssl *ssl = cfg->ssl;
2368 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2369 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2372 if (!reconfiguring) {
2373 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2374 * usage has changed. */
2375 vlan_splinters_changed = false;
2376 if (vlan_splinters_enabled_anywhere) {
2377 HMAP_FOR_EACH (br, node, &all_bridges) {
2378 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2379 vlan_splinters_changed = true;
2385 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2386 idl_seqno = ovsdb_idl_get_seqno(idl);
2388 reconf_txn = ovsdb_idl_txn_create(idl);
2389 bridge_reconfigure(cfg);
2391 /* We still need to reconfigure to avoid dangling pointers to
2392 * now-destroyed ovsrec structures inside bridge data. */
2393 bridge_reconfigure(&null_cfg);
2398 if (reconfiguring) {
2400 reconf_txn = ovsdb_idl_txn_create(idl);
2403 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2404 reconfiguring = false;
2407 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2410 /* If we are completing our initial configuration for this run
2411 * of ovs-vswitchd, then keep the transaction around to monitor
2412 * it for completion. */
2413 if (!initial_config_done) {
2414 initial_config_done = true;
2415 daemonize_txn = reconf_txn;
2422 ovsdb_idl_txn_commit(reconf_txn);
2423 ovsdb_idl_txn_destroy(reconf_txn);
2427 if (daemonize_txn) {
2428 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2429 if (status != TXN_INCOMPLETE) {
2430 ovsdb_idl_txn_destroy(daemonize_txn);
2431 daemonize_txn = NULL;
2433 /* ovs-vswitchd has completed initialization, so allow the
2434 * process that forked us to exit successfully. */
2435 daemonize_complete();
2437 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2441 /* Refresh interface and mirror stats if necessary. */
2442 if (time_msec() >= iface_stats_timer) {
2444 struct ovsdb_idl_txn *txn;
2446 txn = ovsdb_idl_txn_create(idl);
2447 HMAP_FOR_EACH (br, node, &all_bridges) {
2451 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2452 struct iface *iface;
2454 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2455 iface_refresh_stats(iface);
2456 iface_refresh_status(iface);
2460 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2461 mirror_refresh_stats(m);
2465 refresh_controller_status();
2466 ovsdb_idl_txn_commit(txn);
2467 ovsdb_idl_txn_destroy(txn); /* XXX */
2470 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2474 instant_stats_run();
2483 ovsdb_idl_wait(idl);
2484 if (daemonize_txn) {
2485 ovsdb_idl_txn_wait(daemonize_txn);
2488 if (reconfiguring) {
2489 poll_immediate_wake();
2493 ofproto_enumerate_types(&types);
2494 SSET_FOR_EACH (type, &types) {
2495 ofproto_type_wait(type);
2497 sset_destroy(&types);
2499 if (!hmap_is_empty(&all_bridges)) {
2502 HMAP_FOR_EACH (br, node, &all_bridges) {
2503 ofproto_wait(br->ofproto);
2505 poll_timer_wait_until(iface_stats_timer);
2508 system_stats_wait();
2509 instant_stats_wait();
2512 /* Adds some memory usage statistics for bridges into 'usage', for use with
2513 * memory_report(). */
2515 bridge_get_memory_usage(struct simap *usage)
2519 HMAP_FOR_EACH (br, node, &all_bridges) {
2520 ofproto_get_memory_usage(br->ofproto, usage);
2524 /* QoS unixctl user interface functions. */
2526 struct qos_unixctl_show_cbdata {
2528 struct iface *iface;
2532 qos_unixctl_show_cb(unsigned int queue_id,
2533 const struct smap *details,
2536 struct qos_unixctl_show_cbdata *data = aux;
2537 struct ds *ds = data->ds;
2538 struct iface *iface = data->iface;
2539 struct netdev_queue_stats stats;
2540 struct smap_node *node;
2543 ds_put_cstr(ds, "\n");
2545 ds_put_format(ds, "Queue %u:\n", queue_id);
2547 ds_put_cstr(ds, "Default:\n");
2550 SMAP_FOR_EACH (node, details) {
2551 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2554 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2556 if (stats.tx_packets != UINT64_MAX) {
2557 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2560 if (stats.tx_bytes != UINT64_MAX) {
2561 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2564 if (stats.tx_errors != UINT64_MAX) {
2565 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2568 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2569 queue_id, ovs_strerror(error));
2574 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2575 const char *argv[], void *aux OVS_UNUSED)
2577 struct ds ds = DS_EMPTY_INITIALIZER;
2578 struct smap smap = SMAP_INITIALIZER(&smap);
2579 struct iface *iface;
2581 struct smap_node *node;
2582 struct qos_unixctl_show_cbdata data;
2585 iface = iface_find(argv[1]);
2587 unixctl_command_reply_error(conn, "no such interface");
2591 netdev_get_qos(iface->netdev, &type, &smap);
2593 if (*type != '\0') {
2594 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2596 SMAP_FOR_EACH (node, &smap) {
2597 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2602 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2605 ds_put_format(&ds, "failed to dump queues: %s",
2606 ovs_strerror(error));
2608 unixctl_command_reply(conn, ds_cstr(&ds));
2610 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2611 unixctl_command_reply_error(conn, ds_cstr(&ds));
2614 smap_destroy(&smap);
2618 /* Bridge reconfiguration functions. */
2620 bridge_create(const struct ovsrec_bridge *br_cfg)
2624 ovs_assert(!bridge_lookup(br_cfg->name));
2625 br = xzalloc(sizeof *br);
2627 br->name = xstrdup(br_cfg->name);
2628 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2631 /* Derive the default Ethernet address from the bridge's UUID. This should
2632 * be unique and it will be stable between ovs-vswitchd runs. */
2633 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2634 eth_addr_mark_random(br->default_ea);
2636 hmap_init(&br->ports);
2637 hmap_init(&br->ifaces);
2638 hmap_init(&br->iface_by_name);
2639 hmap_init(&br->mirrors);
2641 hmap_init(&br->if_cfg_todo);
2642 list_init(&br->ofpp_garbage);
2644 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2648 bridge_destroy(struct bridge *br)
2651 struct mirror *mirror, *next_mirror;
2652 struct port *port, *next_port;
2653 struct if_cfg *if_cfg, *next_if_cfg;
2654 struct ofpp_garbage *garbage, *next_garbage;
2656 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2659 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2660 mirror_destroy(mirror);
2662 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2663 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2666 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2667 &br->ofpp_garbage) {
2668 list_remove(&garbage->list_node);
2672 hmap_remove(&all_bridges, &br->node);
2673 ofproto_destroy(br->ofproto);
2674 hmap_destroy(&br->ifaces);
2675 hmap_destroy(&br->ports);
2676 hmap_destroy(&br->iface_by_name);
2677 hmap_destroy(&br->mirrors);
2678 hmap_destroy(&br->if_cfg_todo);
2685 static struct bridge *
2686 bridge_lookup(const char *name)
2690 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2691 if (!strcmp(br->name, name)) {
2698 /* Handle requests for a listing of all flows known by the OpenFlow
2699 * stack, including those normally hidden. */
2701 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2702 const char *argv[], void *aux OVS_UNUSED)
2707 br = bridge_lookup(argv[1]);
2709 unixctl_command_reply_error(conn, "Unknown bridge");
2714 ofproto_get_all_flows(br->ofproto, &results);
2716 unixctl_command_reply(conn, ds_cstr(&results));
2717 ds_destroy(&results);
2720 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2721 * connections and reconnect. If BRIDGE is not specified, then all bridges
2722 * drop their controller connections and reconnect. */
2724 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2725 const char *argv[], void *aux OVS_UNUSED)
2729 br = bridge_lookup(argv[1]);
2731 unixctl_command_reply_error(conn, "Unknown bridge");
2734 ofproto_reconnect_controllers(br->ofproto);
2736 HMAP_FOR_EACH (br, node, &all_bridges) {
2737 ofproto_reconnect_controllers(br->ofproto);
2740 unixctl_command_reply(conn, NULL);
2744 bridge_get_controllers(const struct bridge *br,
2745 struct ovsrec_controller ***controllersp)
2747 struct ovsrec_controller **controllers;
2748 size_t n_controllers;
2750 controllers = br->cfg->controller;
2751 n_controllers = br->cfg->n_controller;
2753 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2759 *controllersp = controllers;
2761 return n_controllers;
2765 bridge_queue_if_cfg(struct bridge *br,
2766 const struct ovsrec_interface *cfg,
2767 const struct ovsrec_port *parent)
2769 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2772 if_cfg->parent = parent;
2773 if_cfg->ofport = iface_pick_ofport(cfg);
2774 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2775 hash_string(if_cfg->cfg->name, 0));
2778 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2779 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2780 * 'br' needs to complete its configuration. */
2782 bridge_add_del_ports(struct bridge *br,
2783 const unsigned long int *splinter_vlans)
2785 struct shash_node *port_node;
2786 struct port *port, *next;
2787 struct shash new_ports;
2790 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2792 /* Collect new ports. */
2793 shash_init(&new_ports);
2794 for (i = 0; i < br->cfg->n_ports; i++) {
2795 const char *name = br->cfg->ports[i]->name;
2796 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2797 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2801 if (bridge_get_controllers(br, NULL)
2802 && !shash_find(&new_ports, br->name)) {
2803 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2804 br->name, br->name);
2806 ovsrec_interface_init(&br->synth_local_iface);
2807 ovsrec_port_init(&br->synth_local_port);
2809 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2810 br->synth_local_port.n_interfaces = 1;
2811 br->synth_local_port.name = br->name;
2813 br->synth_local_iface.name = br->name;
2814 br->synth_local_iface.type = "internal";
2816 br->synth_local_ifacep = &br->synth_local_iface;
2818 shash_add(&new_ports, br->name, &br->synth_local_port);
2821 if (splinter_vlans) {
2822 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2825 /* Get rid of deleted ports.
2826 * Get rid of deleted interfaces on ports that still exist. */
2827 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2828 port->cfg = shash_find_data(&new_ports, port->name);
2832 port_del_ifaces(port);
2836 /* Update iface->cfg and iface->type in interfaces that still exist.
2837 * Add new interfaces to creation queue. */
2838 SHASH_FOR_EACH (port_node, &new_ports) {
2839 const struct ovsrec_port *port = port_node->data;
2842 for (i = 0; i < port->n_interfaces; i++) {
2843 const struct ovsrec_interface *cfg = port->interfaces[i];
2844 struct iface *iface = iface_lookup(br, cfg->name);
2845 const char *type = iface_get_type(cfg, br->cfg);
2850 } else if (!strcmp(type, "null")) {
2851 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2852 " may be removed in February 2013. Please email"
2853 " dev@openvswitch.org with concerns.",
2856 bridge_queue_if_cfg(br, cfg, port);
2861 shash_destroy(&new_ports);
2864 /* Initializes 'oc' appropriately as a management service controller for
2867 * The caller must free oc->target when it is no longer needed. */
2869 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2870 struct ofproto_controller *oc)
2872 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2873 oc->max_backoff = 0;
2874 oc->probe_interval = 60;
2875 oc->band = OFPROTO_OUT_OF_BAND;
2877 oc->burst_limit = 0;
2878 oc->enable_async_msgs = true;
2881 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2883 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2884 struct ofproto_controller *oc)
2888 oc->target = c->target;
2889 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2890 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2891 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2892 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2893 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2894 oc->burst_limit = (c->controller_burst_limit
2895 ? *c->controller_burst_limit : 0);
2896 oc->enable_async_msgs = (!c->enable_async_messages
2897 || *c->enable_async_messages);
2898 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2899 if (dscp < 0 || dscp > 63) {
2900 dscp = DSCP_DEFAULT;
2905 /* Configures the IP stack for 'br''s local interface properly according to the
2906 * configuration in 'c'. */
2908 bridge_configure_local_iface_netdev(struct bridge *br,
2909 struct ovsrec_controller *c)
2911 struct netdev *netdev;
2912 struct in_addr mask, gateway;
2914 struct iface *local_iface;
2917 /* If there's no local interface or no IP address, give up. */
2918 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2919 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2923 /* Bring up the local interface. */
2924 netdev = local_iface->netdev;
2925 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2927 /* Configure the IP address and netmask. */
2928 if (!c->local_netmask
2929 || !inet_aton(c->local_netmask, &mask)
2931 mask.s_addr = guess_netmask(ip.s_addr);
2933 if (!netdev_set_in4(netdev, ip, mask)) {
2934 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2935 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2938 /* Configure the default gateway. */
2939 if (c->local_gateway
2940 && inet_aton(c->local_gateway, &gateway)
2941 && gateway.s_addr) {
2942 if (!netdev_add_router(netdev, gateway)) {
2943 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2944 br->name, IP_ARGS(gateway.s_addr));
2949 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2950 * in either string are treated as equal to any number of slashes in the other,
2951 * e.g. "x///y" is equal to "x/y".
2953 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2954 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2955 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2956 * 'b' against a prefix of 'a'.
2959 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2961 const char *b_start = b;
2963 if (b - b_start >= b_stoplen) {
2965 } else if (*a != *b) {
2967 } else if (*a == '/') {
2968 a += strspn(a, "/");
2969 b += strspn(b, "/");
2970 } else if (*a == '\0') {
2980 bridge_configure_remotes(struct bridge *br,
2981 const struct sockaddr_in *managers, size_t n_managers)
2983 bool disable_in_band;
2985 struct ovsrec_controller **controllers;
2986 size_t n_controllers;
2988 enum ofproto_fail_mode fail_mode;
2990 struct ofproto_controller *ocs;
2994 /* Check if we should disable in-band control on this bridge. */
2995 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2998 /* Set OpenFlow queue ID for in-band control. */
2999 ofproto_set_in_band_queue(br->ofproto,
3000 smap_get_int(&br->cfg->other_config,
3001 "in-band-queue", -1));
3003 if (disable_in_band) {
3004 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3006 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3009 n_controllers = bridge_get_controllers(br, &controllers);
3011 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3014 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3015 for (i = 0; i < n_controllers; i++) {
3016 struct ovsrec_controller *c = controllers[i];
3018 if (!strncmp(c->target, "punix:", 6)
3019 || !strncmp(c->target, "unix:", 5)) {
3020 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3023 if (!strncmp(c->target, "unix:", 5)) {
3024 /* Connect to a listening socket */
3025 whitelist = xasprintf("unix:%s/", ovs_rundir());
3026 if (strchr(c->target, '/') &&
3027 !equal_pathnames(c->target, whitelist,
3028 strlen(whitelist))) {
3029 /* Absolute path specified, but not in ovs_rundir */
3030 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3031 "controller \"%s\" due to possibility for "
3032 "remote exploit. Instead, specify socket "
3033 "in whitelisted \"%s\" or connect to "
3034 "\"unix:%s/%s.mgmt\" (which is always "
3035 "available without special configuration).",
3036 br->name, c->target, whitelist,
3037 ovs_rundir(), br->name);
3042 whitelist = xasprintf("punix:%s/%s.controller",
3043 ovs_rundir(), br->name);
3044 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3045 /* Prevent remote ovsdb-server users from accessing
3046 * arbitrary Unix domain sockets and overwriting arbitrary
3048 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3049 "controller \"%s\" due to possibility of "
3050 "overwriting local files. Instead, specify "
3051 "whitelisted \"%s\" or connect to "
3052 "\"unix:%s/%s.mgmt\" (which is always "
3053 "available without special configuration).",
3054 br->name, c->target, whitelist,
3055 ovs_rundir(), br->name);
3064 bridge_configure_local_iface_netdev(br, c);
3065 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3066 if (disable_in_band) {
3067 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3072 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3073 bridge_get_allowed_versions(br));
3074 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3077 /* Set the fail-mode. */
3078 fail_mode = !br->cfg->fail_mode
3079 || !strcmp(br->cfg->fail_mode, "standalone")
3080 ? OFPROTO_FAIL_STANDALONE
3081 : OFPROTO_FAIL_SECURE;
3082 ofproto_set_fail_mode(br->ofproto, fail_mode);
3084 /* Configure OpenFlow controller connection snooping. */
3085 if (!ofproto_has_snoops(br->ofproto)) {
3089 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3090 ovs_rundir(), br->name));
3091 ofproto_set_snoops(br->ofproto, &snoops);
3092 sset_destroy(&snoops);
3097 bridge_configure_tables(struct bridge *br)
3099 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3103 n_tables = ofproto_get_n_tables(br->ofproto);
3105 for (i = 0; i < n_tables; i++) {
3106 struct ofproto_table_settings s;
3109 s.max_flows = UINT_MAX;
3113 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3114 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3117 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3118 s.max_flows = *cfg->flow_limit;
3120 if (cfg->overflow_policy
3121 && !strcmp(cfg->overflow_policy, "evict")) {
3124 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3125 for (k = 0; k < cfg->n_groups; k++) {
3126 const char *string = cfg->groups[k];
3129 msg = mf_parse_subfield__(&s.groups[k], &string);
3131 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3132 "'groups' (%s)", br->name, i, msg);
3134 } else if (*string) {
3135 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3136 "element '%s' contains trailing garbage",
3137 br->name, i, cfg->groups[k]);
3145 ofproto_configure_table(br->ofproto, i, &s);
3149 for (; j < br->cfg->n_flow_tables; j++) {
3150 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3151 "%"PRId64" not supported by this datapath", br->name,
3152 br->cfg->key_flow_tables[j]);
3157 bridge_configure_dp_desc(struct bridge *br)
3159 ofproto_set_dp_desc(br->ofproto,
3160 smap_get(&br->cfg->other_config, "dp-desc"));
3163 /* Port functions. */
3165 static struct port *
3166 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3170 port = xzalloc(sizeof *port);
3172 port->name = xstrdup(cfg->name);
3174 list_init(&port->ifaces);
3176 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3180 /* Deletes interfaces from 'port' that are no longer configured for it. */
3182 port_del_ifaces(struct port *port)
3184 struct iface *iface, *next;
3185 struct sset new_ifaces;
3188 /* Collect list of new interfaces. */
3189 sset_init(&new_ifaces);
3190 for (i = 0; i < port->cfg->n_interfaces; i++) {
3191 const char *name = port->cfg->interfaces[i]->name;
3192 const char *type = port->cfg->interfaces[i]->type;
3193 if (strcmp(type, "null")) {
3194 sset_add(&new_ifaces, name);
3198 /* Get rid of deleted interfaces. */
3199 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3200 if (!sset_contains(&new_ifaces, iface->name)) {
3201 iface_destroy(iface);
3205 sset_destroy(&new_ifaces);
3209 port_destroy(struct port *port)
3212 struct bridge *br = port->bridge;
3213 struct iface *iface, *next;
3216 ofproto_bundle_unregister(br->ofproto, port);
3219 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3220 iface_destroy(iface);
3223 hmap_remove(&br->ports, &port->hmap_node);
3229 static struct port *
3230 port_lookup(const struct bridge *br, const char *name)
3234 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3236 if (!strcmp(port->name, name)) {
3244 enable_lacp(struct port *port, bool *activep)
3246 if (!port->cfg->lacp) {
3247 /* XXX when LACP implementation has been sufficiently tested, enable by
3248 * default and make active on bonded ports. */
3250 } else if (!strcmp(port->cfg->lacp, "off")) {
3252 } else if (!strcmp(port->cfg->lacp, "active")) {
3255 } else if (!strcmp(port->cfg->lacp, "passive")) {
3259 VLOG_WARN("port %s: unknown LACP mode %s",
3260 port->name, port->cfg->lacp);
3265 static struct lacp_settings *
3266 port_configure_lacp(struct port *port, struct lacp_settings *s)
3268 const char *lacp_time, *system_id;
3271 if (!enable_lacp(port, &s->active)) {
3275 s->name = port->name;
3277 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3279 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3280 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3281 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3282 " address.", port->name, system_id);
3286 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3289 if (eth_addr_is_zero(s->id)) {
3290 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3294 /* Prefer bondable links if unspecified. */
3295 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3297 s->priority = (priority > 0 && priority <= UINT16_MAX
3299 : UINT16_MAX - !list_is_short(&port->ifaces));
3301 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3302 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3307 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3309 int priority, portid, key;
3311 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3312 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3314 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3316 if (portid <= 0 || portid > UINT16_MAX) {
3317 portid = ofp_to_u16(iface->ofp_port);
3320 if (priority <= 0 || priority > UINT16_MAX) {
3321 priority = UINT16_MAX;
3324 if (key < 0 || key > UINT16_MAX) {
3328 s->name = iface->name;
3330 s->priority = priority;
3335 port_configure_bond(struct port *port, struct bond_settings *s)
3337 const char *detect_s;
3338 struct iface *iface;
3339 int miimon_interval;
3341 s->name = port->name;
3343 if (port->cfg->bond_mode) {
3344 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3345 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3346 port->name, port->cfg->bond_mode,
3347 bond_mode_to_string(s->balance));
3350 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3352 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3353 * active-backup. At some point we should remove this warning. */
3354 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3355 " in previous versions, the default bond_mode was"
3356 " balance-slb", port->name,
3357 bond_mode_to_string(s->balance));
3359 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3360 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3361 "please use another bond type or disable flood_vlans",
3365 miimon_interval = smap_get_int(&port->cfg->other_config,
3366 "bond-miimon-interval", 0);
3367 if (miimon_interval <= 0) {
3368 miimon_interval = 200;
3371 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3372 if (!detect_s || !strcmp(detect_s, "carrier")) {
3373 miimon_interval = 0;
3374 } else if (strcmp(detect_s, "miimon")) {
3375 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3376 "defaulting to carrier", port->name, detect_s);
3377 miimon_interval = 0;
3380 s->up_delay = MAX(0, port->cfg->bond_updelay);
3381 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3382 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3383 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3384 "bond-rebalance-interval", 10000);
3385 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3386 s->rebalance_interval = 1000;
3389 s->fake_iface = port->cfg->bond_fake_iface;
3391 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3392 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3396 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3397 * instead of obtaining it from the database. */
3399 port_is_synthetic(const struct port *port)
3401 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3404 /* Interface functions. */
3407 iface_is_internal(const struct ovsrec_interface *iface,
3408 const struct ovsrec_bridge *br)
3410 /* The local port and "internal" ports are always "internal". */
3411 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3414 /* Returns the correct network device type for interface 'iface' in bridge
3417 iface_get_type(const struct ovsrec_interface *iface,
3418 const struct ovsrec_bridge *br)
3422 /* The local port always has type "internal". Other ports take
3423 * their type from the database and default to "system" if none is
3425 if (iface_is_internal(iface, br)) {
3428 type = iface->type[0] ? iface->type : "system";
3431 return ofproto_port_open_type(br->datapath_type, type);
3435 iface_destroy(struct iface *iface)
3438 struct port *port = iface->port;
3439 struct bridge *br = port->bridge;
3441 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3442 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3445 if (iface->ofp_port != OFPP_NONE) {
3446 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3449 list_remove(&iface->port_elem);
3450 hmap_remove(&br->iface_by_name, &iface->name_node);
3452 netdev_close(iface->netdev);
3459 static struct iface *
3460 iface_lookup(const struct bridge *br, const char *name)
3462 struct iface *iface;
3464 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3465 &br->iface_by_name) {
3466 if (!strcmp(iface->name, name)) {
3474 static struct iface *
3475 iface_find(const char *name)
3477 const struct bridge *br;
3479 HMAP_FOR_EACH (br, node, &all_bridges) {
3480 struct iface *iface = iface_lookup(br, name);
3489 static struct if_cfg *
3490 if_cfg_lookup(const struct bridge *br, const char *name)
3492 struct if_cfg *if_cfg;
3494 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3496 if (!strcmp(if_cfg->cfg->name, name)) {
3504 static struct iface *
3505 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3507 struct iface *iface;
3509 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3511 if (iface->ofp_port == ofp_port) {
3518 /* Set Ethernet address of 'iface', if one is specified in the configuration
3521 iface_set_mac(struct iface *iface)
3523 uint8_t ea[ETH_ADDR_LEN];
3525 if (!strcmp(iface->type, "internal")
3526 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3527 if (iface->ofp_port == OFPP_LOCAL) {
3528 VLOG_ERR("interface %s: ignoring mac in Interface record "
3529 "(use Bridge record to set local port's mac)",
3531 } else if (eth_addr_is_multicast(ea)) {
3532 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3535 int error = netdev_set_etheraddr(iface->netdev, ea);
3537 VLOG_ERR("interface %s: setting MAC failed (%s)",
3538 iface->name, ovs_strerror(error));
3544 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3546 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3549 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3550 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3551 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3555 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3556 * sets the "ofport" field to -1.
3558 * This is appropriate when 'if_cfg''s interface cannot be created or is
3559 * otherwise invalid. */
3561 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3563 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3564 ovsrec_interface_set_status(if_cfg, NULL);
3565 ovsrec_interface_set_admin_state(if_cfg, NULL);
3566 ovsrec_interface_set_duplex(if_cfg, NULL);
3567 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3568 ovsrec_interface_set_link_state(if_cfg, NULL);
3569 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3570 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3571 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3572 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3573 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3574 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3575 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3579 struct iface_delete_queues_cbdata {
3580 struct netdev *netdev;
3581 const struct ovsdb_datum *queues;
3585 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3587 union ovsdb_atom atom;
3589 atom.integer = target;
3590 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3594 iface_delete_queues(unsigned int queue_id,
3595 const struct smap *details OVS_UNUSED, void *cbdata_)
3597 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3599 if (!queue_ids_include(cbdata->queues, queue_id)) {
3600 netdev_delete_queue(cbdata->netdev, queue_id);
3605 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3607 struct ofpbuf queues_buf;
3609 ofpbuf_init(&queues_buf, 0);
3611 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3612 netdev_set_qos(iface->netdev, NULL, NULL);
3614 struct iface_delete_queues_cbdata cbdata;
3618 /* Configure top-level Qos for 'iface'. */
3619 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3621 /* Deconfigure queues that were deleted. */
3622 cbdata.netdev = iface->netdev;
3623 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3625 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3627 /* Configure queues for 'iface'. */
3629 for (i = 0; i < qos->n_queues; i++) {
3630 const struct ovsrec_queue *queue = qos->value_queues[i];
3631 unsigned int queue_id = qos->key_queues[i];
3633 if (queue_id == 0) {
3637 if (queue->n_dscp == 1) {
3638 struct ofproto_port_queue *port_queue;
3640 port_queue = ofpbuf_put_uninit(&queues_buf,
3641 sizeof *port_queue);
3642 port_queue->queue = queue_id;
3643 port_queue->dscp = queue->dscp[0];
3646 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3649 struct smap details;
3651 smap_init(&details);
3652 netdev_set_queue(iface->netdev, 0, &details);
3653 smap_destroy(&details);
3657 if (iface->ofp_port != OFPP_NONE) {
3658 const struct ofproto_port_queue *port_queues = queues_buf.data;
3659 size_t n_queues = queues_buf.size / sizeof *port_queues;
3661 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3662 port_queues, n_queues);
3665 netdev_set_policing(iface->netdev,
3666 iface->cfg->ingress_policing_rate,
3667 iface->cfg->ingress_policing_burst);
3669 ofpbuf_uninit(&queues_buf);
3673 iface_configure_cfm(struct iface *iface)
3675 const struct ovsrec_interface *cfg = iface->cfg;
3676 const char *opstate_str;
3677 const char *cfm_ccm_vlan;
3678 struct cfm_settings s;
3679 struct smap netdev_args;
3681 if (!cfg->n_cfm_mpid) {
3682 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3686 s.check_tnl_key = false;
3687 smap_init(&netdev_args);
3688 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3689 const char *key = smap_get(&netdev_args, "key");
3690 const char *in_key = smap_get(&netdev_args, "in_key");
3692 s.check_tnl_key = (key && !strcmp(key, "flow"))
3693 || (in_key && !strcmp(in_key, "flow"));
3695 smap_destroy(&netdev_args);
3697 s.mpid = *cfg->cfm_mpid;
3698 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3699 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3700 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3702 if (s.interval <= 0) {
3706 if (!cfm_ccm_vlan) {
3708 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3709 s.ccm_vlan = CFM_RANDOM_VLAN;
3711 s.ccm_vlan = atoi(cfm_ccm_vlan);
3712 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3717 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3719 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3721 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3722 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3724 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3727 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3728 * instead of obtaining it from the database. */
3730 iface_is_synthetic(const struct iface *iface)
3732 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3736 iface_pick_ofport(const struct ovsrec_interface *cfg)
3738 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3740 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3745 /* Port mirroring. */
3747 static struct mirror *
3748 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3752 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3753 if (uuid_equals(uuid, &m->uuid)) {
3761 bridge_configure_mirrors(struct bridge *br)
3763 const struct ovsdb_datum *mc;
3764 unsigned long *flood_vlans;
3765 struct mirror *m, *next;
3768 /* Get rid of deleted mirrors. */
3769 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3770 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3771 union ovsdb_atom atom;
3773 atom.uuid = m->uuid;
3774 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3779 /* Add new mirrors and reconfigure existing ones. */
3780 for (i = 0; i < br->cfg->n_mirrors; i++) {
3781 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3782 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3784 m = mirror_create(br, cfg);
3787 if (!mirror_configure(m)) {
3792 /* Update flooded vlans (for RSPAN). */
3793 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3794 br->cfg->n_flood_vlans);
3795 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3796 bitmap_free(flood_vlans);
3799 static struct mirror *
3800 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3804 m = xzalloc(sizeof *m);
3805 m->uuid = cfg->header_.uuid;
3806 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3808 m->name = xstrdup(cfg->name);
3814 mirror_destroy(struct mirror *m)
3817 struct bridge *br = m->bridge;
3820 ofproto_mirror_unregister(br->ofproto, m);
3823 hmap_remove(&br->mirrors, &m->hmap_node);
3830 mirror_collect_ports(struct mirror *m,
3831 struct ovsrec_port **in_ports, int n_in_ports,
3832 void ***out_portsp, size_t *n_out_portsp)
3834 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3835 size_t n_out_ports = 0;
3838 for (i = 0; i < n_in_ports; i++) {
3839 const char *name = in_ports[i]->name;
3840 struct port *port = port_lookup(m->bridge, name);
3842 out_ports[n_out_ports++] = port;
3844 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3845 "port %s", m->bridge->name, m->name, name);
3848 *out_portsp = out_ports;
3849 *n_out_portsp = n_out_ports;
3853 mirror_configure(struct mirror *m)
3855 const struct ovsrec_mirror *cfg = m->cfg;
3856 struct ofproto_mirror_settings s;
3859 if (strcmp(cfg->name, m->name)) {
3861 m->name = xstrdup(cfg->name);
3865 /* Get output port or VLAN. */
3866 if (cfg->output_port) {
3867 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3868 if (!s.out_bundle) {
3869 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3870 m->bridge->name, m->name);
3873 s.out_vlan = UINT16_MAX;
3875 if (cfg->output_vlan) {
3876 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3877 "output vlan; ignoring output vlan",
3878 m->bridge->name, m->name);
3880 } else if (cfg->output_vlan) {
3881 /* The database should prevent invalid VLAN values. */
3882 s.out_bundle = NULL;
3883 s.out_vlan = *cfg->output_vlan;
3885 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3886 m->bridge->name, m->name);
3890 /* Get port selection. */
3891 if (cfg->select_all) {
3892 size_t n_ports = hmap_count(&m->bridge->ports);
3893 void **ports = xmalloc(n_ports * sizeof *ports);
3898 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3908 /* Get ports, dropping ports that don't exist.
3909 * The IDL ensures that there are no duplicates. */
3910 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3911 &s.srcs, &s.n_srcs);
3912 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3913 &s.dsts, &s.n_dsts);
3916 /* Get VLAN selection. */
3917 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3920 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3923 if (s.srcs != s.dsts) {
3932 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3934 * This is deprecated. It is only for compatibility with broken device drivers
3935 * in old versions of Linux that do not properly support VLANs when VLAN
3936 * devices are not used. When broken device drivers are no longer in
3937 * widespread use, we will delete these interfaces. */
3939 static struct ovsrec_port **recs;
3940 static size_t n_recs, allocated_recs;
3942 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3943 * splinters are reconfigured. */
3945 register_rec(struct ovsrec_port *rec)
3947 if (n_recs >= allocated_recs) {
3948 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3950 recs[n_recs++] = rec;
3953 /* Frees all of the ports registered with register_reg(). */
3955 free_registered_recs(void)
3959 for (i = 0; i < n_recs; i++) {
3960 struct ovsrec_port *port = recs[i];
3963 for (j = 0; j < port->n_interfaces; j++) {
3964 struct ovsrec_interface *iface = port->interfaces[j];
3969 smap_destroy(&port->other_config);
3970 free(port->interfaces);
3978 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3981 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3983 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
3987 /* Figures out the set of VLANs that are in use for the purpose of VLAN
3990 * If VLAN splinters are enabled on at least one interface and any VLANs are in
3991 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
3992 * 4095 will not be set). The caller is responsible for freeing the bitmap,
3995 * If VLANs splinters are not enabled on any interface or if no VLANs are in
3996 * use, returns NULL.
3998 * Updates 'vlan_splinters_enabled_anywhere'. */
3999 static unsigned long int *
4000 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4002 unsigned long int *splinter_vlans;
4003 struct sset splinter_ifaces;
4004 const char *real_dev_name;
4005 struct shash *real_devs;
4006 struct shash_node *node;
4010 /* Free space allocated for synthesized ports and interfaces, since we're
4011 * in the process of reconstructing all of them. */
4012 free_registered_recs();
4014 splinter_vlans = bitmap_allocate(4096);
4015 sset_init(&splinter_ifaces);
4016 vlan_splinters_enabled_anywhere = false;
4017 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4018 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4021 for (j = 0; j < br_cfg->n_ports; j++) {
4022 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4025 for (k = 0; k < port_cfg->n_interfaces; k++) {
4026 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4028 if (vlan_splinters_is_enabled(iface_cfg)) {
4029 vlan_splinters_enabled_anywhere = true;
4030 sset_add(&splinter_ifaces, iface_cfg->name);
4031 vlan_bitmap_from_array__(port_cfg->trunks,
4037 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4038 bitmap_set1(splinter_vlans, *port_cfg->tag);
4043 if (!vlan_splinters_enabled_anywhere) {
4044 free(splinter_vlans);
4045 sset_destroy(&splinter_ifaces);
4049 HMAP_FOR_EACH (br, node, &all_bridges) {
4051 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4055 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4056 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4057 * device to be created for it. */
4058 bitmap_set0(splinter_vlans, 0);
4059 bitmap_set0(splinter_vlans, 4095);
4061 /* Delete all VLAN devices that we don't need. */
4063 real_devs = vlandev_get_real_devs();
4064 SHASH_FOR_EACH (node, real_devs) {
4065 const struct vlan_real_dev *real_dev = node->data;
4066 const struct vlan_dev *vlan_dev;
4067 bool real_dev_has_splinters;
4069 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4071 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4072 if (!real_dev_has_splinters
4073 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4074 struct netdev *netdev;
4076 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4077 if (!netdev_get_in4(netdev, NULL, NULL) ||
4078 !netdev_get_in6(netdev, NULL)) {
4079 vlandev_del(vlan_dev->name);
4081 /* It has an IP address configured, so we don't own
4082 * it. Don't delete it. */
4084 netdev_close(netdev);
4091 /* Add all VLAN devices that we need. */
4092 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4095 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4096 if (!vlandev_get_name(real_dev_name, vid)) {
4097 vlandev_add(real_dev_name, vid);
4104 sset_destroy(&splinter_ifaces);
4106 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4107 free(splinter_vlans);
4110 return splinter_vlans;
4113 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4116 configure_splinter_port(struct port *port)
4118 struct ofproto *ofproto = port->bridge->ofproto;
4119 ofp_port_t realdev_ofp_port;
4120 const char *realdev_name;
4121 struct iface *vlandev, *realdev;
4123 ofproto_bundle_unregister(port->bridge->ofproto, port);
4125 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4128 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4129 realdev = iface_lookup(port->bridge, realdev_name);
4130 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4132 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4136 static struct ovsrec_port *
4137 synthesize_splinter_port(const char *real_dev_name,
4138 const char *vlan_dev_name, int vid)
4140 struct ovsrec_interface *iface;
4141 struct ovsrec_port *port;
4143 iface = xmalloc(sizeof *iface);
4144 ovsrec_interface_init(iface);
4145 iface->name = xstrdup(vlan_dev_name);
4146 iface->type = "system";
4148 port = xmalloc(sizeof *port);
4149 ovsrec_port_init(port);
4150 port->interfaces = xmemdup(&iface, sizeof iface);
4151 port->n_interfaces = 1;
4152 port->name = xstrdup(vlan_dev_name);
4153 port->vlan_mode = "splinter";
4154 port->tag = xmalloc(sizeof *port->tag);
4157 smap_add(&port->other_config, "realdev", real_dev_name);
4163 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4164 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4165 * 1-bit in the 'splinter_vlans' bitmap. */
4167 add_vlan_splinter_ports(struct bridge *br,
4168 const unsigned long int *splinter_vlans,
4169 struct shash *ports)
4173 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4174 * we're modifying 'ports'. */
4175 for (i = 0; i < br->cfg->n_ports; i++) {
4176 const char *name = br->cfg->ports[i]->name;
4177 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4180 for (j = 0; j < port_cfg->n_interfaces; j++) {
4181 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4183 if (vlan_splinters_is_enabled(iface_cfg)) {
4184 const char *real_dev_name;
4187 real_dev_name = iface_cfg->name;
4188 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4189 const char *vlan_dev_name;
4191 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4193 && !shash_find(ports, vlan_dev_name)) {
4194 shash_add(ports, vlan_dev_name,
4195 synthesize_splinter_port(
4196 real_dev_name, vlan_dev_name, vid));
4205 mirror_refresh_stats(struct mirror *m)
4207 struct ofproto *ofproto = m->bridge->ofproto;
4208 uint64_t tx_packets, tx_bytes;
4211 size_t stat_cnt = 0;
4213 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4214 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4218 if (tx_packets != UINT64_MAX) {
4219 keys[stat_cnt] = "tx_packets";
4220 values[stat_cnt] = tx_packets;
4223 if (tx_bytes != UINT64_MAX) {
4224 keys[stat_cnt] = "tx_bytes";
4225 values[stat_cnt] = tx_bytes;
4229 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);