1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/ofproto.h"
43 #include "poll-loop.h"
47 #include "socket-util.h"
49 #include "stream-ssl.h"
51 #include "system-stats.h"
56 #include "lib/vswitch-idl.h"
57 #include "xenserver.h"
59 #include "sflow_api.h"
60 #include "vlan-bitmap.h"
62 VLOG_DEFINE_THIS_MODULE(bridge);
64 COVERAGE_DEFINE(bridge_reconfigure);
66 /* Configuration of an uninstantiated iface. */
68 struct hmap_node hmap_node; /* Node in bridge's if_cfg_todo. */
69 const struct ovsrec_interface *cfg; /* Interface record. */
70 const struct ovsrec_port *parent; /* Parent port record. */
71 ofp_port_t ofport; /* Requested OpenFlow port number. */
74 /* OpenFlow port slated for removal from ofproto. */
76 struct list list_node; /* Node in bridge's ofpp_garbage. */
77 ofp_port_t ofp_port; /* Port to be deleted. */
81 /* These members are always valid. */
82 struct list port_elem; /* Element in struct port's "ifaces" list. */
83 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
87 /* These members are valid only after bridge_reconfigure() causes them to
89 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
90 ofp_port_t ofp_port; /* OpenFlow port number, */
91 /* OFPP_NONE if unknown. */
92 struct netdev *netdev; /* Network device. */
93 const char *type; /* Usually same as cfg->type. */
94 const struct ovsrec_interface *cfg;
98 struct uuid uuid; /* UUID of this "mirror" record in database. */
99 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
100 struct bridge *bridge;
102 const struct ovsrec_mirror *cfg;
106 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
107 struct bridge *bridge;
110 const struct ovsrec_port *cfg;
112 /* An ordinary bridge port has 1 interface.
113 * A bridge port for bonding has at least 2 interfaces. */
114 struct list ifaces; /* List of "struct iface"s. */
118 struct hmap_node node; /* In 'all_bridges'. */
119 char *name; /* User-specified arbitrary name. */
120 char *type; /* Datapath type. */
121 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
122 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
123 const struct ovsrec_bridge *cfg;
125 /* OpenFlow switch processing. */
126 struct ofproto *ofproto; /* OpenFlow switch. */
129 struct hmap ports; /* "struct port"s indexed by name. */
130 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
131 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
133 struct list ofpp_garbage; /* "struct ofpp_garbage" slated for removal. */
134 struct hmap if_cfg_todo; /* "struct if_cfg"s slated for creation.
135 Indexed on 'cfg->name'. */
137 /* Port mirroring. */
138 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
146 /* All bridges, indexed by name. */
147 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
149 /* OVSDB IDL used to obtain configuration. */
150 static struct ovsdb_idl *idl;
152 /* We want to complete daemonization, fully detaching from our parent process,
153 * only after we have completed our initial configuration, committed our state
154 * to the database, and received confirmation back from the database server
155 * that it applied the commit. This allows our parent process to know that,
156 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
157 * to have already been filled in. (It is only "very likely" rather than
158 * certain because there is always a slim possibility that the transaction will
159 * fail or that some other client has added new bridges, ports, etc. while
160 * ovs-vswitchd was configuring using an old configuration.)
162 * We only need to do this once for our initial configuration at startup, so
163 * 'initial_config_done' tracks whether we've already done it. While we are
164 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
165 * itself and is otherwise NULL. */
166 static bool initial_config_done;
167 static struct ovsdb_idl_txn *daemonize_txn;
169 /* Most recently processed IDL sequence number. */
170 static unsigned int idl_seqno;
172 /* Each time this timer expires, the bridge fetches interface and mirror
173 * statistics and pushes them into the database. */
174 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
175 static long long int iface_stats_timer = LLONG_MIN;
177 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
178 * expensive. This can cause bridge_reconfigure() to take a long time during
179 * which no other work can be done. To deal with this problem, we limit port
180 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
181 * call to bridge_reconfigure(). If there is more work to do after the limit
182 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
183 * This allows the rest of the code to catch up on important things like
184 * forwarding packets. */
185 #define OFP_PORT_ACTION_WINDOW 10
186 static bool reconfiguring = false;
188 static void add_del_bridges(const struct ovsrec_open_vswitch *);
189 static void bridge_update_ofprotos(void);
190 static void bridge_create(const struct ovsrec_bridge *);
191 static void bridge_destroy(struct bridge *);
192 static struct bridge *bridge_lookup(const char *name);
193 static unixctl_cb_func bridge_unixctl_dump_flows;
194 static unixctl_cb_func bridge_unixctl_reconnect;
195 static size_t bridge_get_controllers(const struct bridge *br,
196 struct ovsrec_controller ***controllersp);
197 static void bridge_add_del_ports(struct bridge *,
198 const unsigned long int *splinter_vlans);
199 static void bridge_refresh_ofp_port(struct bridge *);
200 static void bridge_configure_flow_miss_model(const char *opt);
201 static void bridge_configure_datapath_id(struct bridge *);
202 static void bridge_configure_netflow(struct bridge *);
203 static void bridge_configure_forward_bpdu(struct bridge *);
204 static void bridge_configure_mac_table(struct bridge *);
205 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
206 static void bridge_configure_ipfix(struct bridge *);
207 static void bridge_configure_stp(struct bridge *);
208 static void bridge_configure_tables(struct bridge *);
209 static void bridge_configure_dp_desc(struct bridge *);
210 static void bridge_configure_remotes(struct bridge *,
211 const struct sockaddr_in *managers,
213 static void bridge_pick_local_hw_addr(struct bridge *,
214 uint8_t ea[ETH_ADDR_LEN],
215 struct iface **hw_addr_iface);
216 static uint64_t bridge_pick_datapath_id(struct bridge *,
217 const uint8_t bridge_ea[ETH_ADDR_LEN],
218 struct iface *hw_addr_iface);
219 static void bridge_queue_if_cfg(struct bridge *,
220 const struct ovsrec_interface *,
221 const struct ovsrec_port *);
222 static uint64_t dpid_from_hash(const void *, size_t nbytes);
223 static bool bridge_has_bond_fake_iface(const struct bridge *,
225 static bool port_is_bond_fake_iface(const struct port *);
227 static unixctl_cb_func qos_unixctl_show;
229 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
230 static void port_del_ifaces(struct port *);
231 static void port_destroy(struct port *);
232 static struct port *port_lookup(const struct bridge *, const char *name);
233 static void port_configure(struct port *);
234 static struct lacp_settings *port_configure_lacp(struct port *,
235 struct lacp_settings *);
236 static void port_configure_bond(struct port *, struct bond_settings *);
237 static bool port_is_synthetic(const struct port *);
239 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
240 static void run_system_stats(void);
242 static void bridge_configure_mirrors(struct bridge *);
243 static struct mirror *mirror_create(struct bridge *,
244 const struct ovsrec_mirror *);
245 static void mirror_destroy(struct mirror *);
246 static bool mirror_configure(struct mirror *);
247 static void mirror_refresh_stats(struct mirror *);
249 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
250 static bool iface_create(struct bridge *, struct if_cfg *,
251 ofp_port_t ofp_port);
252 static bool iface_is_internal(const struct ovsrec_interface *iface,
253 const struct ovsrec_bridge *br);
254 static const char *iface_get_type(const struct ovsrec_interface *,
255 const struct ovsrec_bridge *);
256 static void iface_destroy(struct iface *);
257 static struct iface *iface_lookup(const struct bridge *, const char *name);
258 static struct iface *iface_find(const char *name);
259 static struct if_cfg *if_cfg_lookup(const struct bridge *, const char *name);
260 static struct iface *iface_from_ofp_port(const struct bridge *,
261 ofp_port_t ofp_port);
262 static void iface_set_mac(struct iface *);
263 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
264 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
265 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
266 static void iface_configure_cfm(struct iface *);
267 static void iface_refresh_cfm_stats(struct iface *);
268 static void iface_refresh_stats(struct iface *);
269 static void iface_refresh_status(struct iface *);
270 static bool iface_is_synthetic(const struct iface *);
271 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
275 * This is deprecated. It is only for compatibility with broken device drivers
276 * in old versions of Linux that do not properly support VLANs when VLAN
277 * devices are not used. When broken device drivers are no longer in
278 * widespread use, we will delete these interfaces. */
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere;
283 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
284 static unsigned long int *collect_splinter_vlans(
285 const struct ovsrec_open_vswitch *);
286 static void configure_splinter_port(struct port *);
287 static void add_vlan_splinter_ports(struct bridge *,
288 const unsigned long int *splinter_vlans,
289 struct shash *ports);
292 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
294 struct shash iface_hints;
295 static bool initialized = false;
302 shash_init(&iface_hints);
305 for (i = 0; i < cfg->n_bridges; i++) {
306 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
309 for (j = 0; j < br_cfg->n_ports; j++) {
310 struct ovsrec_port *port_cfg = br_cfg->ports[j];
313 for (k = 0; k < port_cfg->n_interfaces; k++) {
314 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
315 struct iface_hint *iface_hint;
317 iface_hint = xmalloc(sizeof *iface_hint);
318 iface_hint->br_name = br_cfg->name;
319 iface_hint->br_type = br_cfg->datapath_type;
320 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
322 shash_add(&iface_hints, if_cfg->name, iface_hint);
328 ofproto_init(&iface_hints);
330 shash_destroy_free_data(&iface_hints);
334 /* Public functions. */
336 /* Initializes the bridge module, configuring it to obtain its configuration
337 * from an OVSDB server accessed over 'remote', which should be a string in a
338 * form acceptable to ovsdb_idl_create(). */
340 bridge_init(const char *remote)
342 /* Create connection to database. */
343 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
344 idl_seqno = ovsdb_idl_get_seqno(idl);
345 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
346 ovsdb_idl_verify_write_only(idl);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
349 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
354 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
357 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
358 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
361 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
363 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
383 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
387 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
388 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
395 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
397 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
406 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
408 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
410 /* Register unixctl commands. */
411 unixctl_command_register("qos/show", "interface", 1, 1,
412 qos_unixctl_show, NULL);
413 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
414 bridge_unixctl_dump_flows, NULL);
415 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
416 bridge_unixctl_reconnect, NULL);
426 struct bridge *br, *next_br;
428 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
431 ovsdb_idl_destroy(idl);
434 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
435 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
436 * responsible for freeing '*managersp' (with free()).
438 * You may be asking yourself "why does ovs-vswitchd care?", because
439 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
440 * should not be and in fact is not directly involved in that. But
441 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
442 * it has to tell in-band control where the managers are to enable that.
443 * (Thus, only managers connected in-band are collected.)
446 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
447 struct sockaddr_in **managersp, size_t *n_managersp)
449 struct sockaddr_in *managers = NULL;
450 size_t n_managers = 0;
454 /* Collect all of the potential targets from the "targets" columns of the
455 * rows pointed to by "manager_options", excluding any that are
458 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
459 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
461 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
462 sset_find_and_delete(&targets, m->target);
464 sset_add(&targets, m->target);
468 /* Now extract the targets' IP addresses. */
469 if (!sset_is_empty(&targets)) {
472 managers = xmalloc(sset_count(&targets) * sizeof *managers);
473 SSET_FOR_EACH (target, &targets) {
474 struct sockaddr_in *sin = &managers[n_managers];
476 if (stream_parse_target_with_default_ports(target,
484 sset_destroy(&targets);
486 *managersp = managers;
487 *n_managersp = n_managers;
491 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
493 unsigned long int *splinter_vlans;
496 COVERAGE_INC(bridge_reconfigure);
498 ovs_assert(!reconfiguring);
499 reconfiguring = true;
501 ofproto_set_flow_eviction_threshold(
502 smap_get_int(&ovs_cfg->other_config, "flow-eviction-threshold",
503 OFPROTO_FLOW_EVICTION_THRESHOLD_DEFAULT));
505 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
506 "force-miss-model"));
508 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
509 * to 'ovs_cfg' while update the "if_cfg_queue", with only very minimal
510 * configuration otherwise.
512 * This is mostly an update to bridge data structures. Nothing is pushed
513 * down to ofproto or lower layers. */
514 add_del_bridges(ovs_cfg);
515 splinter_vlans = collect_splinter_vlans(ovs_cfg);
516 HMAP_FOR_EACH (br, node, &all_bridges) {
517 bridge_add_del_ports(br, splinter_vlans);
519 free(splinter_vlans);
521 /* Delete datapaths that are no longer configured, and create ones which
522 * don't exist but should. */
523 bridge_update_ofprotos();
525 /* Make sure each "struct iface" has a correct ofp_port in its ofproto. */
526 HMAP_FOR_EACH (br, node, &all_bridges) {
527 bridge_refresh_ofp_port(br);
530 /* Clear database records for "if_cfg"s which haven't been instantiated. */
531 HMAP_FOR_EACH (br, node, &all_bridges) {
532 struct if_cfg *if_cfg;
534 HMAP_FOR_EACH (if_cfg, hmap_node, &br->if_cfg_todo) {
535 iface_clear_db_record(if_cfg->cfg);
539 reconfigure_system_stats(ovs_cfg);
543 bridge_reconfigure_ofp(void)
545 long long int deadline;
549 deadline = time_msec() + OFP_PORT_ACTION_WINDOW;
551 /* The kernel will reject any attempt to add a given port to a datapath if
552 * that port already belongs to a different datapath, so we must do all
553 * port deletions before any port additions. */
554 HMAP_FOR_EACH (br, node, &all_bridges) {
555 struct ofpp_garbage *garbage, *next;
557 LIST_FOR_EACH_SAFE (garbage, next, list_node, &br->ofpp_garbage) {
558 /* It's a bit dangerous to call bridge_run_fast() here as ofproto's
559 * internal datastructures may not be consistent. Eventually, when
560 * port additions and deletions are cheaper, these calls should be
563 ofproto_port_del(br->ofproto, garbage->ofp_port);
564 list_remove(&garbage->list_node);
568 if (time_msec() >= deadline) {
575 HMAP_FOR_EACH (br, node, &all_bridges) {
576 struct if_cfg *if_cfg, *next;
578 HMAP_FOR_EACH_SAFE (if_cfg, next, hmap_node, &br->if_cfg_todo) {
579 iface_create(br, if_cfg, OFPP_NONE);
581 if (time_msec() >= deadline) {
591 bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
593 struct sockaddr_in *managers;
594 int sflow_bridge_number;
599 ovs_assert(reconfiguring);
600 done = bridge_reconfigure_ofp();
602 /* Complete the configuration. */
603 sflow_bridge_number = 0;
604 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
605 HMAP_FOR_EACH (br, node, &all_bridges) {
608 /* We need the datapath ID early to allow LACP ports to use it as the
609 * default system ID. */
610 bridge_configure_datapath_id(br);
612 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
615 port_configure(port);
617 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
618 iface_configure_cfm(iface);
619 iface_configure_qos(iface, port->cfg->qos);
620 iface_set_mac(iface);
621 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
625 bridge_configure_mirrors(br);
626 bridge_configure_forward_bpdu(br);
627 bridge_configure_mac_table(br);
628 bridge_configure_remotes(br, managers, n_managers);
629 bridge_configure_netflow(br);
630 bridge_configure_sflow(br, &sflow_bridge_number);
631 bridge_configure_ipfix(br);
632 bridge_configure_stp(br);
633 bridge_configure_tables(br);
634 bridge_configure_dp_desc(br);
636 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
637 /* XXX: Remove this warning message eventually. */
638 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
639 " moved to the Open_vSwitch table. Ignoring its"
640 " setting in the bridge table.");
648 /* Delete ofprotos which aren't configured or have the wrong type. Create
649 * ofprotos which don't exist but need to. */
651 bridge_update_ofprotos(void)
653 struct bridge *br, *next;
658 /* Delete ofprotos with no bridge or with the wrong type. */
661 ofproto_enumerate_types(&types);
662 SSET_FOR_EACH (type, &types) {
665 ofproto_enumerate_names(type, &names);
666 SSET_FOR_EACH (name, &names) {
667 br = bridge_lookup(name);
668 if (!br || strcmp(type, br->type)) {
669 ofproto_delete(name, type);
673 sset_destroy(&names);
674 sset_destroy(&types);
676 /* Add ofprotos for bridges which don't have one yet. */
677 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
685 /* Remove ports from any datapath with the same name as 'br'. If we
686 * don't do this, creating 'br''s ofproto will fail because a port with
687 * the same name as its local port already exists. */
688 HMAP_FOR_EACH (br2, node, &all_bridges) {
689 struct ofproto_port ofproto_port;
695 if (!ofproto_port_query_by_name(br2->ofproto, br->name,
697 error = ofproto_port_del(br2->ofproto, ofproto_port.ofp_port);
699 VLOG_ERR("failed to delete port %s: %s", ofproto_port.name,
700 ovs_strerror(error));
702 ofproto_port_destroy(&ofproto_port);
706 error = ofproto_create(br->name, br->type, &br->ofproto);
708 VLOG_ERR("failed to create bridge %s: %s", br->name,
709 ovs_strerror(error));
716 port_configure(struct port *port)
718 const struct ovsrec_port *cfg = port->cfg;
719 struct bond_settings bond_settings;
720 struct lacp_settings lacp_settings;
721 struct ofproto_bundle_settings s;
724 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
725 configure_splinter_port(port);
734 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
735 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
736 s.slaves[s.n_slaves++] = iface->ofp_port;
741 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
745 /* Get VLAN trunks. */
748 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
752 if (cfg->vlan_mode) {
753 if (!strcmp(cfg->vlan_mode, "access")) {
754 s.vlan_mode = PORT_VLAN_ACCESS;
755 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
756 s.vlan_mode = PORT_VLAN_TRUNK;
757 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
758 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
759 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
760 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
762 /* This "can't happen" because ovsdb-server should prevent it. */
763 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
764 s.vlan_mode = PORT_VLAN_TRUNK;
768 s.vlan_mode = PORT_VLAN_ACCESS;
770 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
774 s.vlan_mode = PORT_VLAN_TRUNK;
777 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
780 /* Get LACP settings. */
781 s.lacp = port_configure_lacp(port, &lacp_settings);
785 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
786 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
787 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
790 s.lacp_slaves = NULL;
793 /* Get bond settings. */
794 if (s.n_slaves > 1) {
795 s.bond = &bond_settings;
796 port_configure_bond(port, &bond_settings);
799 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
800 netdev_set_miimon_interval(iface->netdev, 0);
805 ofproto_bundle_register(port->bridge->ofproto, port, &s);
814 bridge_configure_flow_miss_model(const char *opt)
816 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
819 if (strcmp(opt, "with-facets")) {
820 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
821 } else if (strcmp(opt, "without-facets")) {
822 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
826 ofproto_set_flow_miss_model(model);
829 /* Pick local port hardware address and datapath ID for 'br'. */
831 bridge_configure_datapath_id(struct bridge *br)
833 uint8_t ea[ETH_ADDR_LEN];
835 struct iface *local_iface;
836 struct iface *hw_addr_iface;
839 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
840 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
842 int error = netdev_set_etheraddr(local_iface->netdev, ea);
844 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
845 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
846 "Ethernet address: %s",
847 br->name, ovs_strerror(error));
850 memcpy(br->ea, ea, ETH_ADDR_LEN);
852 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
853 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
854 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
855 ofproto_set_datapath_id(br->ofproto, dpid);
858 dpid_string = xasprintf("%016"PRIx64, dpid);
859 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
863 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
866 bridge_get_allowed_versions(struct bridge *br)
868 if (!br->cfg->n_protocols)
871 return ofputil_versions_from_strings(br->cfg->protocols,
872 br->cfg->n_protocols);
875 /* Set NetFlow configuration on 'br'. */
877 bridge_configure_netflow(struct bridge *br)
879 struct ovsrec_netflow *cfg = br->cfg->netflow;
880 struct netflow_options opts;
883 ofproto_set_netflow(br->ofproto, NULL);
887 memset(&opts, 0, sizeof opts);
889 /* Get default NetFlow configuration from datapath.
890 * Apply overrides from 'cfg'. */
891 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
892 if (cfg->engine_type) {
893 opts.engine_type = *cfg->engine_type;
895 if (cfg->engine_id) {
896 opts.engine_id = *cfg->engine_id;
899 /* Configure active timeout interval. */
900 opts.active_timeout = cfg->active_timeout;
901 if (!opts.active_timeout) {
902 opts.active_timeout = -1;
903 } else if (opts.active_timeout < 0) {
904 VLOG_WARN("bridge %s: active timeout interval set to negative "
905 "value, using default instead (%d seconds)", br->name,
906 NF_ACTIVE_TIMEOUT_DEFAULT);
907 opts.active_timeout = -1;
910 /* Add engine ID to interface number to disambiguate bridgs? */
911 opts.add_id_to_iface = cfg->add_id_to_interface;
912 if (opts.add_id_to_iface) {
913 if (opts.engine_id > 0x7f) {
914 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
915 "another vswitch, choose an engine id less than 128",
918 if (hmap_count(&br->ports) > 508) {
919 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
920 "another port when more than 508 ports are used",
926 sset_init(&opts.collectors);
927 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
930 if (ofproto_set_netflow(br->ofproto, &opts)) {
931 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
933 sset_destroy(&opts.collectors);
936 /* Set sFlow configuration on 'br'. */
938 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
940 const struct ovsrec_sflow *cfg = br->cfg->sflow;
941 struct ovsrec_controller **controllers;
942 struct ofproto_sflow_options oso;
943 size_t n_controllers;
947 ofproto_set_sflow(br->ofproto, NULL);
951 memset(&oso, 0, sizeof oso);
953 sset_init(&oso.targets);
954 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
956 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
958 oso.sampling_rate = *cfg->sampling;
961 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
963 oso.polling_interval = *cfg->polling;
966 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
968 oso.header_len = *cfg->header;
971 oso.sub_id = (*sflow_bridge_number)++;
972 oso.agent_device = cfg->agent;
974 oso.control_ip = NULL;
975 n_controllers = bridge_get_controllers(br, &controllers);
976 for (i = 0; i < n_controllers; i++) {
977 if (controllers[i]->local_ip) {
978 oso.control_ip = controllers[i]->local_ip;
982 ofproto_set_sflow(br->ofproto, &oso);
984 sset_destroy(&oso.targets);
987 /* Set IPFIX configuration on 'br'. */
989 bridge_configure_ipfix(struct bridge *br)
991 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
992 const struct ovsrec_flow_sample_collector_set *fe_cfg;
993 struct ofproto_ipfix_bridge_exporter_options be_opts;
994 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
995 size_t n_fe_opts = 0;
997 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
998 if (fe_cfg->bridge == br->cfg) {
1003 if (!be_cfg && n_fe_opts == 0) {
1004 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1009 memset(&be_opts, 0, sizeof be_opts);
1011 sset_init(&be_opts.targets);
1012 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1014 if (be_cfg->sampling) {
1015 be_opts.sampling_rate = *be_cfg->sampling;
1017 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1019 if (be_cfg->obs_domain_id) {
1020 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1022 if (be_cfg->obs_point_id) {
1023 be_opts.obs_point_id = *be_cfg->obs_point_id;
1027 if (n_fe_opts > 0) {
1028 struct ofproto_ipfix_flow_exporter_options *opts;
1029 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1031 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1032 if (fe_cfg->bridge == br->cfg) {
1033 opts->collector_set_id = fe_cfg->id;
1034 sset_init(&opts->targets);
1035 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1036 fe_cfg->ipfix->n_targets);
1042 ofproto_set_ipfix(br->ofproto, be_cfg ? &be_opts : NULL, fe_opts,
1046 sset_destroy(&be_opts.targets);
1049 if (n_fe_opts > 0) {
1050 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1052 for (i = 0; i < n_fe_opts; i++) {
1053 sset_destroy(&opts->targets);
1061 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1062 struct ofproto_port_stp_settings *port_s,
1063 int *port_num_counter, unsigned long *port_num_bitmap)
1065 const char *config_str;
1066 struct iface *iface;
1068 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1069 port_s->enable = false;
1072 port_s->enable = true;
1075 /* STP over bonds is not supported. */
1076 if (!list_is_singleton(&port->ifaces)) {
1077 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1079 port_s->enable = false;
1083 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1085 /* Internal ports shouldn't participate in spanning tree, so
1087 if (!strcmp(iface->type, "internal")) {
1088 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1089 port_s->enable = false;
1093 /* STP on mirror output ports is not supported. */
1094 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1095 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1096 port_s->enable = false;
1100 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1102 unsigned long int port_num = strtoul(config_str, NULL, 0);
1103 int port_idx = port_num - 1;
1105 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1106 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1107 port_s->enable = false;
1111 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1112 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1113 port->name, port_num);
1114 port_s->enable = false;
1117 bitmap_set1(port_num_bitmap, port_idx);
1118 port_s->port_num = port_idx;
1120 if (*port_num_counter >= STP_MAX_PORTS) {
1121 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1122 port_s->enable = false;
1126 port_s->port_num = (*port_num_counter)++;
1129 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1131 port_s->path_cost = strtoul(config_str, NULL, 10);
1133 enum netdev_features current;
1136 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1137 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1138 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1141 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1143 port_s->priority = strtoul(config_str, NULL, 0);
1145 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1149 /* Set spanning tree configuration on 'br'. */
1151 bridge_configure_stp(struct bridge *br)
1153 if (!br->cfg->stp_enable) {
1154 ofproto_set_stp(br->ofproto, NULL);
1156 struct ofproto_stp_settings br_s;
1157 const char *config_str;
1159 int port_num_counter;
1160 unsigned long *port_num_bitmap;
1162 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1164 uint8_t ea[ETH_ADDR_LEN];
1166 if (eth_addr_from_string(config_str, ea)) {
1167 br_s.system_id = eth_addr_to_uint64(ea);
1169 br_s.system_id = eth_addr_to_uint64(br->ea);
1170 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1171 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1174 br_s.system_id = eth_addr_to_uint64(br->ea);
1177 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1179 br_s.priority = strtoul(config_str, NULL, 0);
1181 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1184 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1186 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1188 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1191 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1193 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1195 br_s.max_age = STP_DEFAULT_MAX_AGE;
1198 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1200 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1202 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1205 /* Configure STP on the bridge. */
1206 if (ofproto_set_stp(br->ofproto, &br_s)) {
1207 VLOG_ERR("bridge %s: could not enable STP", br->name);
1211 /* Users must either set the port number with the "stp-port-num"
1212 * configuration on all ports or none. If manual configuration
1213 * is not done, then we allocate them sequentially. */
1214 port_num_counter = 0;
1215 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1216 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1217 struct ofproto_port_stp_settings port_s;
1218 struct iface *iface;
1220 port_configure_stp(br->ofproto, port, &port_s,
1221 &port_num_counter, port_num_bitmap);
1223 /* As bonds are not supported, just apply configuration to
1224 * all interfaces. */
1225 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1226 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1228 VLOG_ERR("port %s: could not enable STP", port->name);
1234 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1235 && port_num_counter) {
1236 VLOG_ERR("bridge %s: must manually configure all STP port "
1237 "IDs or none, disabling", br->name);
1238 ofproto_set_stp(br->ofproto, NULL);
1240 bitmap_free(port_num_bitmap);
1245 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1247 const struct port *port = port_lookup(br, name);
1248 return port && port_is_bond_fake_iface(port);
1252 port_is_bond_fake_iface(const struct port *port)
1254 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1258 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1260 struct bridge *br, *next;
1261 struct shash new_br;
1264 /* Collect new bridges' names and types. */
1265 shash_init(&new_br);
1266 for (i = 0; i < cfg->n_bridges; i++) {
1267 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1268 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1270 if (strchr(br_cfg->name, '/')) {
1271 /* Prevent remote ovsdb-server users from accessing arbitrary
1272 * directories, e.g. consider a bridge named "../../../etc/". */
1273 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1275 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1276 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1280 /* Get rid of deleted bridges or those whose types have changed.
1281 * Update 'cfg' of bridges that still exist. */
1282 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1283 br->cfg = shash_find_data(&new_br, br->name);
1284 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1285 br->cfg->datapath_type))) {
1290 /* Add new bridges. */
1291 for (i = 0; i < cfg->n_bridges; i++) {
1292 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1293 struct bridge *br = bridge_lookup(br_cfg->name);
1295 bridge_create(br_cfg);
1299 shash_destroy(&new_br);
1303 iface_set_ofp_port(struct iface *iface, ofp_port_t ofp_port)
1305 struct bridge *br = iface->port->bridge;
1307 ovs_assert(iface->ofp_port == OFPP_NONE && ofp_port != OFPP_NONE);
1308 iface->ofp_port = ofp_port;
1309 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1310 hash_ofp_port(ofp_port));
1311 iface_set_ofport(iface->cfg, ofp_port);
1314 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1315 * Returns 0 if successful, otherwise a positive errno value. */
1317 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1318 struct netdev *netdev)
1322 error = netdev_set_config(netdev, &iface_cfg->options);
1324 VLOG_WARN("could not configure network device %s (%s)",
1325 iface_cfg->name, ovs_strerror(error));
1330 /* This function determines whether 'ofproto_port', which is attached to
1331 * br->ofproto's datapath, is one that we want in 'br'.
1333 * If it is, it returns true, after creating an iface (if necessary),
1334 * configuring the iface's netdev according to the iface's options, and setting
1335 * iface's ofp_port member to 'ofproto_port->ofp_port'.
1337 * If, on the other hand, 'port' should be removed, it returns false. The
1338 * caller should later detach the port from br->ofproto. */
1340 bridge_refresh_one_ofp_port(struct bridge *br,
1341 const struct ofproto_port *ofproto_port)
1343 const char *name = ofproto_port->name;
1344 const char *type = ofproto_port->type;
1345 ofp_port_t ofp_port = ofproto_port->ofp_port;
1347 struct iface *iface = iface_lookup(br, name);
1349 /* Check that the name-to-number mapping is one-to-one. */
1350 if (iface->ofp_port != OFPP_NONE) {
1351 VLOG_WARN("bridge %s: interface %s reported twice",
1354 } else if (iface_from_ofp_port(br, ofp_port)) {
1355 VLOG_WARN("bridge %s: interface %"PRIu16" reported twice",
1356 br->name, ofp_port);
1360 /* There's a configured interface named 'name'. */
1361 if (strcmp(type, iface->type)
1362 || iface_set_netdev_config(iface->cfg, iface->netdev)) {
1363 /* It's the wrong type, or it's the right type but can't be
1364 * configured as the user requested, so we must destroy it. */
1367 /* It's the right type and configured correctly. Keep it. */
1368 iface_set_ofp_port(iface, ofp_port);
1371 } else if (bridge_has_bond_fake_iface(br, name)
1372 && !strcmp(type, "internal")) {
1373 /* It's a bond fake iface. Keep it. */
1376 /* There's no configured interface named 'name', but there might be an
1377 * interface of that name queued to be created.
1379 * If there is, and it has the correct type, then try to configure it
1380 * and add it. If that's successful, we'll keep it. Otherwise, we'll
1381 * delete it and later try to re-add it. */
1382 struct if_cfg *if_cfg = if_cfg_lookup(br, name);
1384 && !strcmp(type, iface_get_type(if_cfg->cfg, br->cfg))
1385 && iface_create(br, if_cfg, ofp_port));
1389 /* Update bridges "if_cfg"s, "struct port"s, and "struct iface"s to be
1390 * consistent with the ofp_ports in "br->ofproto". */
1392 bridge_refresh_ofp_port(struct bridge *br)
1394 struct ofproto_port_dump dump;
1395 struct ofproto_port ofproto_port;
1396 struct port *port, *port_next;
1398 /* Clear each "struct iface"s ofp_port so we can get its correct value. */
1399 hmap_clear(&br->ifaces);
1400 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1401 struct iface *iface;
1403 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1404 iface->ofp_port = OFPP_NONE;
1408 /* Obtain the correct "ofp_port"s from ofproto. Find any if_cfg's which
1409 * already exist in the datapath and promote them to full fledged "struct
1410 * iface"s. Mark ports in the datapath which don't belong as garbage. */
1411 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
1412 if (!bridge_refresh_one_ofp_port(br, &ofproto_port)) {
1413 struct ofpp_garbage *garbage = xmalloc(sizeof *garbage);
1414 garbage->ofp_port = ofproto_port.ofp_port;
1415 list_push_front(&br->ofpp_garbage, &garbage->list_node);
1419 /* Some ifaces may not have "ofp_port"s in ofproto and therefore don't
1420 * deserve to have "struct iface"s. Demote these to "if_cfg"s so that
1421 * later they can be added to ofproto. */
1422 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
1423 struct iface *iface, *iface_next;
1425 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
1426 if (iface->ofp_port == OFPP_NONE) {
1427 bridge_queue_if_cfg(br, iface->cfg, port->cfg);
1428 iface_destroy(iface);
1432 if (list_is_empty(&port->ifaces)) {
1438 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1439 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1440 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1443 * If successful, returns 0 and stores the network device in '*netdevp'. On
1444 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1446 iface_do_create(const struct bridge *br,
1447 const struct if_cfg *if_cfg,
1448 ofp_port_t *ofp_portp, struct netdev **netdevp)
1450 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1451 const struct ovsrec_port *port_cfg = if_cfg->parent;
1452 struct netdev *netdev = NULL;
1455 if (netdev_is_reserved_name(iface_cfg->name)) {
1456 VLOG_WARN("could not create interface %s, name is reserved",
1462 error = netdev_open(iface_cfg->name,
1463 iface_get_type(iface_cfg, br->cfg), &netdev);
1465 VLOG_WARN("could not open network device %s (%s)",
1466 iface_cfg->name, ovs_strerror(error));
1470 error = iface_set_netdev_config(iface_cfg, netdev);
1475 if (*ofp_portp == OFPP_NONE) {
1476 ofp_port_t ofp_port = if_cfg->ofport;
1478 error = ofproto_port_add(br->ofproto, netdev, &ofp_port);
1482 *ofp_portp = ofp_port;
1484 VLOG_INFO("bridge %s: added interface %s on port %d",
1485 br->name, iface_cfg->name, *ofp_portp);
1487 VLOG_DBG("bridge %s: interface %s is on port %d",
1488 br->name, iface_cfg->name, *ofp_portp);
1491 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1492 || iface_is_internal(iface_cfg, br->cfg)) {
1493 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1501 netdev_close(netdev);
1505 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1506 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1507 * automatically allocated for the iface. Takes ownership of and
1508 * deallocates 'if_cfg'.
1510 * Return true if an iface is successfully created, false otherwise. */
1512 iface_create(struct bridge *br, struct if_cfg *if_cfg, ofp_port_t ofp_port)
1514 const struct ovsrec_interface *iface_cfg = if_cfg->cfg;
1515 const struct ovsrec_port *port_cfg = if_cfg->parent;
1517 struct netdev *netdev;
1518 struct iface *iface;
1523 /* Do the bits that can fail up front.
1525 * It's a bit dangerous to call bridge_run_fast() here as ofproto's
1526 * internal datastructures may not be consistent. Eventually, when port
1527 * additions and deletions are cheaper, these calls should be removed. */
1529 ovs_assert(!iface_lookup(br, iface_cfg->name));
1530 error = iface_do_create(br, if_cfg, &ofp_port, &netdev);
1533 iface_set_ofport(iface_cfg, OFPP_NONE);
1534 iface_clear_db_record(iface_cfg);
1539 /* Get or create the port structure. */
1540 port = port_lookup(br, port_cfg->name);
1542 port = port_create(br, port_cfg);
1545 /* Create the iface structure. */
1546 iface = xzalloc(sizeof *iface);
1547 list_push_back(&port->ifaces, &iface->port_elem);
1548 hmap_insert(&br->iface_by_name, &iface->name_node,
1549 hash_string(iface_cfg->name, 0));
1551 iface->name = xstrdup(iface_cfg->name);
1552 iface->ofp_port = OFPP_NONE;
1553 iface->netdev = netdev;
1554 iface->type = iface_get_type(iface_cfg, br->cfg);
1555 iface->cfg = iface_cfg;
1557 iface_set_ofp_port(iface, ofp_port);
1559 /* Populate initial status in database. */
1560 iface_refresh_stats(iface);
1561 iface_refresh_status(iface);
1563 /* Add bond fake iface if necessary. */
1564 if (port_is_bond_fake_iface(port)) {
1565 struct ofproto_port ofproto_port;
1567 if (ofproto_port_query_by_name(br->ofproto, port->name,
1569 struct netdev *netdev;
1572 error = netdev_open(port->name, "internal", &netdev);
1574 ofp_port_t fake_ofp_port = if_cfg->ofport;
1576 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1577 netdev_close(netdev);
1579 VLOG_WARN("could not open network device %s (%s)",
1580 port->name, ovs_strerror(error));
1583 /* Already exists, nothing to do. */
1584 ofproto_port_destroy(&ofproto_port);
1589 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
1595 /* Set forward BPDU option. */
1597 bridge_configure_forward_bpdu(struct bridge *br)
1599 ofproto_set_forward_bpdu(br->ofproto,
1600 smap_get_bool(&br->cfg->other_config,
1605 /* Set MAC learning table configuration for 'br'. */
1607 bridge_configure_mac_table(struct bridge *br)
1609 const char *idle_time_str;
1612 const char *mac_table_size_str;
1615 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1616 idle_time = (idle_time_str && atoi(idle_time_str)
1617 ? atoi(idle_time_str)
1618 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1620 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1621 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1622 ? atoi(mac_table_size_str)
1625 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1629 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1630 struct iface **hw_addr_iface)
1632 struct hmapx mirror_output_ports;
1635 bool found_addr = false;
1639 *hw_addr_iface = NULL;
1641 /* Did the user request a particular MAC? */
1642 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1643 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1644 if (eth_addr_is_multicast(ea)) {
1645 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1646 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1647 } else if (eth_addr_is_zero(ea)) {
1648 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1654 /* Mirror output ports don't participate in picking the local hardware
1655 * address. ofproto can't help us find out whether a given port is a
1656 * mirror output because we haven't configured mirrors yet, so we need to
1657 * accumulate them ourselves. */
1658 hmapx_init(&mirror_output_ports);
1659 for (i = 0; i < br->cfg->n_mirrors; i++) {
1660 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1661 if (m->output_port) {
1662 hmapx_add(&mirror_output_ports, m->output_port);
1666 /* Otherwise choose the minimum non-local MAC address among all of the
1668 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1669 uint8_t iface_ea[ETH_ADDR_LEN];
1670 struct iface *candidate;
1671 struct iface *iface;
1673 /* Mirror output ports don't participate. */
1674 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1678 /* Choose the MAC address to represent the port. */
1680 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1681 /* Find the interface with this Ethernet address (if any) so that
1682 * we can provide the correct devname to the caller. */
1683 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1684 uint8_t candidate_ea[ETH_ADDR_LEN];
1685 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1686 && eth_addr_equals(iface_ea, candidate_ea)) {
1691 /* Choose the interface whose MAC address will represent the port.
1692 * The Linux kernel bonding code always chooses the MAC address of
1693 * the first slave added to a bond, and the Fedora networking
1694 * scripts always add slaves to a bond in alphabetical order, so
1695 * for compatibility we choose the interface with the name that is
1696 * first in alphabetical order. */
1697 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1698 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1703 /* The local port doesn't count (since we're trying to choose its
1704 * MAC address anyway). */
1705 if (iface->ofp_port == OFPP_LOCAL) {
1710 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1716 /* Compare against our current choice. */
1717 if (!eth_addr_is_multicast(iface_ea) &&
1718 !eth_addr_is_local(iface_ea) &&
1719 !eth_addr_is_reserved(iface_ea) &&
1720 !eth_addr_is_zero(iface_ea) &&
1721 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1723 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1724 *hw_addr_iface = iface;
1730 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1731 *hw_addr_iface = NULL;
1734 hmapx_destroy(&mirror_output_ports);
1737 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1738 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1739 * an interface on 'br', then that interface must be passed in as
1740 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1741 * 'hw_addr_iface' must be passed in as a null pointer. */
1743 bridge_pick_datapath_id(struct bridge *br,
1744 const uint8_t bridge_ea[ETH_ADDR_LEN],
1745 struct iface *hw_addr_iface)
1748 * The procedure for choosing a bridge MAC address will, in the most
1749 * ordinary case, also choose a unique MAC that we can use as a datapath
1750 * ID. In some special cases, though, multiple bridges will end up with
1751 * the same MAC address. This is OK for the bridges, but it will confuse
1752 * the OpenFlow controller, because each datapath needs a unique datapath
1755 * Datapath IDs must be unique. It is also very desirable that they be
1756 * stable from one run to the next, so that policy set on a datapath
1759 const char *datapath_id;
1762 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1763 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1767 if (!hw_addr_iface) {
1769 * A purely internal bridge, that is, one that has no non-virtual
1770 * network devices on it at all, is difficult because it has no
1771 * natural unique identifier at all.
1773 * When the host is a XenServer, we handle this case by hashing the
1774 * host's UUID with the name of the bridge. Names of bridges are
1775 * persistent across XenServer reboots, although they can be reused if
1776 * an internal network is destroyed and then a new one is later
1777 * created, so this is fairly effective.
1779 * When the host is not a XenServer, we punt by using a random MAC
1780 * address on each run.
1782 const char *host_uuid = xenserver_get_host_uuid();
1784 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1785 dpid = dpid_from_hash(combined, strlen(combined));
1791 return eth_addr_to_uint64(bridge_ea);
1795 dpid_from_hash(const void *data, size_t n)
1797 uint8_t hash[SHA1_DIGEST_SIZE];
1799 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1800 sha1_bytes(data, n, hash);
1801 eth_addr_mark_random(hash);
1802 return eth_addr_to_uint64(hash);
1806 iface_refresh_status(struct iface *iface)
1810 enum netdev_features current;
1814 uint8_t mac[ETH_ADDR_LEN];
1818 if (iface_is_synthetic(iface)) {
1824 if (!netdev_get_status(iface->netdev, &smap)) {
1825 ovsrec_interface_set_status(iface->cfg, &smap);
1827 ovsrec_interface_set_status(iface->cfg, NULL);
1830 smap_destroy(&smap);
1832 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1833 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1835 ovsrec_interface_set_duplex(iface->cfg,
1836 netdev_features_is_full_duplex(current)
1838 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1840 ovsrec_interface_set_duplex(iface->cfg, NULL);
1841 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1844 error = netdev_get_mtu(iface->netdev, &mtu);
1847 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1849 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1852 error = netdev_get_etheraddr(iface->netdev, mac);
1854 char mac_string[32];
1856 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1857 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1859 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1862 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1863 * if there is no valid ifindex number. */
1864 ifindex64 = netdev_get_ifindex(iface->netdev);
1865 if (ifindex64 < 0) {
1868 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1871 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1874 iface_refresh_cfm_stats(struct iface *iface)
1876 const struct ovsrec_interface *cfg = iface->cfg;
1877 struct ofproto_cfm_status status;
1879 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1880 iface->ofp_port, &status)) {
1881 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1882 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1883 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1884 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1885 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1887 const char *reasons[CFM_FAULT_N_REASONS];
1888 int64_t cfm_health = status.health;
1889 bool faulted = status.faults != 0;
1892 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1895 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1896 int reason = 1 << i;
1897 if (status.faults & reason) {
1898 reasons[j++] = cfm_fault_reason_to_str(reason);
1901 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1903 if (status.remote_opstate >= 0) {
1904 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1905 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1907 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1910 ovsrec_interface_set_cfm_remote_mpids(cfg,
1911 (const int64_t *)status.rmps,
1913 if (cfm_health >= 0) {
1914 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1916 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1924 iface_refresh_stats(struct iface *iface)
1926 #define IFACE_STATS \
1927 IFACE_STAT(rx_packets, "rx_packets") \
1928 IFACE_STAT(tx_packets, "tx_packets") \
1929 IFACE_STAT(rx_bytes, "rx_bytes") \
1930 IFACE_STAT(tx_bytes, "tx_bytes") \
1931 IFACE_STAT(rx_dropped, "rx_dropped") \
1932 IFACE_STAT(tx_dropped, "tx_dropped") \
1933 IFACE_STAT(rx_errors, "rx_errors") \
1934 IFACE_STAT(tx_errors, "tx_errors") \
1935 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1936 IFACE_STAT(rx_over_errors, "rx_over_err") \
1937 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1938 IFACE_STAT(collisions, "collisions")
1940 #define IFACE_STAT(MEMBER, NAME) + 1
1941 enum { N_IFACE_STATS = IFACE_STATS };
1943 int64_t values[N_IFACE_STATS];
1944 char *keys[N_IFACE_STATS];
1947 struct netdev_stats stats;
1949 if (iface_is_synthetic(iface)) {
1953 /* Intentionally ignore return value, since errors will set 'stats' to
1954 * all-1s, and we will deal with that correctly below. */
1955 netdev_get_stats(iface->netdev, &stats);
1957 /* Copy statistics into keys[] and values[]. */
1959 #define IFACE_STAT(MEMBER, NAME) \
1960 if (stats.MEMBER != UINT64_MAX) { \
1962 values[n] = stats.MEMBER; \
1967 ovs_assert(n <= N_IFACE_STATS);
1969 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1974 br_refresh_stp_status(struct bridge *br)
1976 struct smap smap = SMAP_INITIALIZER(&smap);
1977 struct ofproto *ofproto = br->ofproto;
1978 struct ofproto_stp_status status;
1980 if (ofproto_get_stp_status(ofproto, &status)) {
1984 if (!status.enabled) {
1985 ovsrec_bridge_set_status(br->cfg, NULL);
1989 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1990 STP_ID_ARGS(status.bridge_id));
1991 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1992 STP_ID_ARGS(status.designated_root));
1993 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1995 ovsrec_bridge_set_status(br->cfg, &smap);
1996 smap_destroy(&smap);
2000 port_refresh_stp_status(struct port *port)
2002 struct ofproto *ofproto = port->bridge->ofproto;
2003 struct iface *iface;
2004 struct ofproto_port_stp_status status;
2006 int64_t int_values[3];
2009 if (port_is_synthetic(port)) {
2013 /* STP doesn't currently support bonds. */
2014 if (!list_is_singleton(&port->ifaces)) {
2015 ovsrec_port_set_status(port->cfg, NULL);
2019 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2021 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2025 if (!status.enabled) {
2026 ovsrec_port_set_status(port->cfg, NULL);
2027 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2031 /* Set Status column. */
2033 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2034 smap_add(&smap, "stp_state", stp_state_name(status.state));
2035 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2036 smap_add(&smap, "stp_role", stp_role_name(status.role));
2037 ovsrec_port_set_status(port->cfg, &smap);
2038 smap_destroy(&smap);
2040 /* Set Statistics column. */
2041 keys[0] = "stp_tx_count";
2042 int_values[0] = status.tx_count;
2043 keys[1] = "stp_rx_count";
2044 int_values[1] = status.rx_count;
2045 keys[2] = "stp_error_count";
2046 int_values[2] = status.error_count;
2048 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2049 ARRAY_SIZE(int_values));
2053 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2055 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2059 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2061 bool enable = enable_system_stats(cfg);
2063 system_stats_enable(enable);
2065 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2070 run_system_stats(void)
2072 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2075 stats = system_stats_run();
2077 struct ovsdb_idl_txn *txn;
2078 struct ovsdb_datum datum;
2080 txn = ovsdb_idl_txn_create(idl);
2081 ovsdb_datum_from_smap(&datum, stats);
2082 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2084 ovsdb_idl_txn_commit(txn);
2085 ovsdb_idl_txn_destroy(txn);
2091 static inline const char *
2092 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2095 case OFPCR12_ROLE_EQUAL:
2097 case OFPCR12_ROLE_MASTER:
2099 case OFPCR12_ROLE_SLAVE:
2101 case OFPCR12_ROLE_NOCHANGE:
2103 return "*** INVALID ROLE ***";
2108 refresh_controller_status(void)
2112 const struct ovsrec_controller *cfg;
2116 /* Accumulate status for controllers on all bridges. */
2117 HMAP_FOR_EACH (br, node, &all_bridges) {
2118 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2121 /* Update each controller in the database with current status. */
2122 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2123 struct ofproto_controller_info *cinfo =
2124 shash_find_data(&info, cfg->target);
2127 struct smap smap = SMAP_INITIALIZER(&smap);
2128 const char **values = cinfo->pairs.values;
2129 const char **keys = cinfo->pairs.keys;
2132 for (i = 0; i < cinfo->pairs.n; i++) {
2133 smap_add(&smap, keys[i], values[i]);
2136 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2137 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2139 ovsrec_controller_set_status(cfg, &smap);
2140 smap_destroy(&smap);
2142 ovsrec_controller_set_is_connected(cfg, false);
2143 ovsrec_controller_set_role(cfg, NULL);
2144 ovsrec_controller_set_status(cfg, NULL);
2148 ofproto_free_ofproto_controller_info(&info);
2153 * Some information in the database must be kept as up-to-date as possible to
2154 * allow controllers to respond rapidly to network outages. We call these
2155 * statistics "instant" stats.
2157 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2158 * assuming that they've changed. The only means we have to determine whether
2159 * they have changed are:
2161 * - Try to commit changes to the database. If nothing changed, then
2162 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2165 * - instant_stats_run() is called late in the run loop, after anything that
2166 * might change any of the instant stats.
2168 * We use these two facts together to avoid waking the process up every
2169 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2172 /* Minimum interval between writing updates to the instant stats to the
2174 #define INSTANT_INTERVAL_MSEC 100
2176 /* Current instant stats database transaction, NULL if there is no ongoing
2178 static struct ovsdb_idl_txn *instant_txn;
2180 /* Next time (in msec on monotonic clock) at which we will update the instant
2182 static long long int instant_next_txn = LLONG_MIN;
2184 /* True if the run loop has run since we last saw that the instant stats were
2185 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2186 * to refresh the instant stats. */
2187 static bool instant_stats_could_have_changed;
2190 instant_stats_run(void)
2192 enum ovsdb_idl_txn_status status;
2194 instant_stats_could_have_changed = true;
2199 if (time_msec() < instant_next_txn) {
2202 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2204 instant_txn = ovsdb_idl_txn_create(idl);
2205 HMAP_FOR_EACH (br, node, &all_bridges) {
2206 struct iface *iface;
2209 br_refresh_stp_status(br);
2211 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2212 port_refresh_stp_status(port);
2215 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2216 enum netdev_flags flags;
2218 const char *link_state;
2219 int64_t link_resets;
2222 if (iface_is_synthetic(iface)) {
2226 current = ofproto_port_is_lacp_current(br->ofproto,
2230 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2232 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2235 error = netdev_get_flags(iface->netdev, &flags);
2237 const char *state = flags & NETDEV_UP ? "up" : "down";
2238 ovsrec_interface_set_admin_state(iface->cfg, state);
2240 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2243 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2244 ovsrec_interface_set_link_state(iface->cfg, link_state);
2246 link_resets = netdev_get_carrier_resets(iface->netdev);
2247 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2249 iface_refresh_cfm_stats(iface);
2252 if (!ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2254 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2255 smap_destroy(&smap);
2261 status = ovsdb_idl_txn_commit(instant_txn);
2262 if (status != TXN_INCOMPLETE) {
2263 ovsdb_idl_txn_destroy(instant_txn);
2266 if (status == TXN_UNCHANGED) {
2267 instant_stats_could_have_changed = false;
2272 instant_stats_wait(void)
2275 ovsdb_idl_txn_wait(instant_txn);
2276 } else if (instant_stats_could_have_changed) {
2277 poll_timer_wait_until(instant_next_txn);
2281 /* Performs periodic activity required by bridges that needs to be done with
2282 * the least possible latency.
2284 * It makes sense to call this function a couple of times per poll loop, to
2285 * provide a significant performance boost on some benchmarks with ofprotos
2286 * that use the ofproto-dpif implementation. */
2288 bridge_run_fast(void)
2295 ofproto_enumerate_types(&types);
2296 SSET_FOR_EACH (type, &types) {
2297 ofproto_type_run_fast(type);
2299 sset_destroy(&types);
2301 HMAP_FOR_EACH (br, node, &all_bridges) {
2302 ofproto_run_fast(br->ofproto);
2309 static struct ovsrec_open_vswitch null_cfg;
2310 const struct ovsrec_open_vswitch *cfg;
2311 struct ovsdb_idl_txn *reconf_txn = NULL;
2315 bool vlan_splinters_changed;
2318 ovsrec_open_vswitch_init(&null_cfg);
2320 /* (Re)configure if necessary. */
2321 if (!reconfiguring) {
2324 if (ovsdb_idl_is_lock_contended(idl)) {
2325 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2326 struct bridge *br, *next_br;
2328 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2329 "disabling this process (pid %ld) until it goes away",
2330 (long int) getpid());
2332 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2335 /* Since we will not be running system_stats_run() in this process
2336 * with the current situation of multiple ovs-vswitchd daemons,
2337 * disable system stats collection. */
2338 system_stats_enable(false);
2340 } else if (!ovsdb_idl_has_lock(idl)) {
2344 cfg = ovsrec_open_vswitch_first(idl);
2346 /* Initialize the ofproto library. This only needs to run once, but
2347 * it must be done after the configuration is set. If the
2348 * initialization has already occurred, bridge_init_ofproto()
2349 * returns immediately. */
2350 bridge_init_ofproto(cfg);
2352 /* Once the value of flow-restore-wait is false, we no longer should
2353 * check its value from the database. */
2354 if (cfg && ofproto_get_flow_restore_wait()) {
2355 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2356 "flow-restore-wait", false));
2359 /* Let each datapath type do the work that it needs to do. */
2361 ofproto_enumerate_types(&types);
2362 SSET_FOR_EACH (type, &types) {
2363 ofproto_type_run(type);
2365 sset_destroy(&types);
2367 /* Let each bridge do the work that it needs to do. */
2368 HMAP_FOR_EACH (br, node, &all_bridges) {
2369 ofproto_run(br->ofproto);
2372 /* Re-configure SSL. We do this on every trip through the main loop,
2373 * instead of just when the database changes, because the contents of the
2374 * key and certificate files can change without the database changing.
2376 * We do this before bridge_reconfigure() because that function might
2377 * initiate SSL connections and thus requires SSL to be configured. */
2378 if (cfg && cfg->ssl) {
2379 const struct ovsrec_ssl *ssl = cfg->ssl;
2381 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2382 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2385 if (!reconfiguring) {
2386 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2387 * usage has changed. */
2388 vlan_splinters_changed = false;
2389 if (vlan_splinters_enabled_anywhere) {
2390 HMAP_FOR_EACH (br, node, &all_bridges) {
2391 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2392 vlan_splinters_changed = true;
2398 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2399 idl_seqno = ovsdb_idl_get_seqno(idl);
2401 reconf_txn = ovsdb_idl_txn_create(idl);
2402 bridge_reconfigure(cfg);
2404 /* We still need to reconfigure to avoid dangling pointers to
2405 * now-destroyed ovsrec structures inside bridge data. */
2406 bridge_reconfigure(&null_cfg);
2411 if (reconfiguring) {
2413 reconf_txn = ovsdb_idl_txn_create(idl);
2416 if (bridge_reconfigure_continue(cfg ? cfg : &null_cfg)) {
2417 reconfiguring = false;
2420 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2423 /* If we are completing our initial configuration for this run
2424 * of ovs-vswitchd, then keep the transaction around to monitor
2425 * it for completion. */
2426 if (!initial_config_done) {
2427 initial_config_done = true;
2428 daemonize_txn = reconf_txn;
2435 ovsdb_idl_txn_commit(reconf_txn);
2436 ovsdb_idl_txn_destroy(reconf_txn);
2440 if (daemonize_txn) {
2441 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2442 if (status != TXN_INCOMPLETE) {
2443 ovsdb_idl_txn_destroy(daemonize_txn);
2444 daemonize_txn = NULL;
2446 /* ovs-vswitchd has completed initialization, so allow the
2447 * process that forked us to exit successfully. */
2448 daemonize_complete();
2450 vlog_enable_async();
2452 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2456 /* Refresh interface and mirror stats if necessary. */
2457 if (time_msec() >= iface_stats_timer) {
2459 struct ovsdb_idl_txn *txn;
2461 txn = ovsdb_idl_txn_create(idl);
2462 HMAP_FOR_EACH (br, node, &all_bridges) {
2466 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2467 struct iface *iface;
2469 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2470 iface_refresh_stats(iface);
2471 iface_refresh_status(iface);
2475 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2476 mirror_refresh_stats(m);
2480 refresh_controller_status();
2481 ovsdb_idl_txn_commit(txn);
2482 ovsdb_idl_txn_destroy(txn); /* XXX */
2485 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2489 instant_stats_run();
2498 ovsdb_idl_wait(idl);
2499 if (daemonize_txn) {
2500 ovsdb_idl_txn_wait(daemonize_txn);
2503 if (reconfiguring) {
2504 poll_immediate_wake();
2508 ofproto_enumerate_types(&types);
2509 SSET_FOR_EACH (type, &types) {
2510 ofproto_type_wait(type);
2512 sset_destroy(&types);
2514 if (!hmap_is_empty(&all_bridges)) {
2517 HMAP_FOR_EACH (br, node, &all_bridges) {
2518 ofproto_wait(br->ofproto);
2520 poll_timer_wait_until(iface_stats_timer);
2523 system_stats_wait();
2524 instant_stats_wait();
2527 /* Adds some memory usage statistics for bridges into 'usage', for use with
2528 * memory_report(). */
2530 bridge_get_memory_usage(struct simap *usage)
2534 HMAP_FOR_EACH (br, node, &all_bridges) {
2535 ofproto_get_memory_usage(br->ofproto, usage);
2539 /* QoS unixctl user interface functions. */
2541 struct qos_unixctl_show_cbdata {
2543 struct iface *iface;
2547 qos_unixctl_show_cb(unsigned int queue_id,
2548 const struct smap *details,
2551 struct qos_unixctl_show_cbdata *data = aux;
2552 struct ds *ds = data->ds;
2553 struct iface *iface = data->iface;
2554 struct netdev_queue_stats stats;
2555 struct smap_node *node;
2558 ds_put_cstr(ds, "\n");
2560 ds_put_format(ds, "Queue %u:\n", queue_id);
2562 ds_put_cstr(ds, "Default:\n");
2565 SMAP_FOR_EACH (node, details) {
2566 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2569 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2571 if (stats.tx_packets != UINT64_MAX) {
2572 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2575 if (stats.tx_bytes != UINT64_MAX) {
2576 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2579 if (stats.tx_errors != UINT64_MAX) {
2580 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2583 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2584 queue_id, ovs_strerror(error));
2589 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2590 const char *argv[], void *aux OVS_UNUSED)
2592 struct ds ds = DS_EMPTY_INITIALIZER;
2593 struct smap smap = SMAP_INITIALIZER(&smap);
2594 struct iface *iface;
2596 struct smap_node *node;
2597 struct qos_unixctl_show_cbdata data;
2600 iface = iface_find(argv[1]);
2602 unixctl_command_reply_error(conn, "no such interface");
2606 netdev_get_qos(iface->netdev, &type, &smap);
2608 if (*type != '\0') {
2609 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2611 SMAP_FOR_EACH (node, &smap) {
2612 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2617 error = netdev_dump_queues(iface->netdev, qos_unixctl_show_cb, &data);
2620 ds_put_format(&ds, "failed to dump queues: %s",
2621 ovs_strerror(error));
2623 unixctl_command_reply(conn, ds_cstr(&ds));
2625 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2626 unixctl_command_reply_error(conn, ds_cstr(&ds));
2629 smap_destroy(&smap);
2633 /* Bridge reconfiguration functions. */
2635 bridge_create(const struct ovsrec_bridge *br_cfg)
2639 ovs_assert(!bridge_lookup(br_cfg->name));
2640 br = xzalloc(sizeof *br);
2642 br->name = xstrdup(br_cfg->name);
2643 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2646 /* Derive the default Ethernet address from the bridge's UUID. This should
2647 * be unique and it will be stable between ovs-vswitchd runs. */
2648 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2649 eth_addr_mark_random(br->default_ea);
2651 hmap_init(&br->ports);
2652 hmap_init(&br->ifaces);
2653 hmap_init(&br->iface_by_name);
2654 hmap_init(&br->mirrors);
2656 hmap_init(&br->if_cfg_todo);
2657 list_init(&br->ofpp_garbage);
2659 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2663 bridge_destroy(struct bridge *br)
2666 struct mirror *mirror, *next_mirror;
2667 struct port *port, *next_port;
2668 struct if_cfg *if_cfg, *next_if_cfg;
2669 struct ofpp_garbage *garbage, *next_garbage;
2671 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2674 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2675 mirror_destroy(mirror);
2677 HMAP_FOR_EACH_SAFE (if_cfg, next_if_cfg, hmap_node, &br->if_cfg_todo) {
2678 hmap_remove(&br->if_cfg_todo, &if_cfg->hmap_node);
2681 LIST_FOR_EACH_SAFE (garbage, next_garbage, list_node,
2682 &br->ofpp_garbage) {
2683 list_remove(&garbage->list_node);
2687 hmap_remove(&all_bridges, &br->node);
2688 ofproto_destroy(br->ofproto);
2689 hmap_destroy(&br->ifaces);
2690 hmap_destroy(&br->ports);
2691 hmap_destroy(&br->iface_by_name);
2692 hmap_destroy(&br->mirrors);
2693 hmap_destroy(&br->if_cfg_todo);
2700 static struct bridge *
2701 bridge_lookup(const char *name)
2705 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2706 if (!strcmp(br->name, name)) {
2713 /* Handle requests for a listing of all flows known by the OpenFlow
2714 * stack, including those normally hidden. */
2716 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2717 const char *argv[], void *aux OVS_UNUSED)
2722 br = bridge_lookup(argv[1]);
2724 unixctl_command_reply_error(conn, "Unknown bridge");
2729 ofproto_get_all_flows(br->ofproto, &results);
2731 unixctl_command_reply(conn, ds_cstr(&results));
2732 ds_destroy(&results);
2735 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2736 * connections and reconnect. If BRIDGE is not specified, then all bridges
2737 * drop their controller connections and reconnect. */
2739 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2740 const char *argv[], void *aux OVS_UNUSED)
2744 br = bridge_lookup(argv[1]);
2746 unixctl_command_reply_error(conn, "Unknown bridge");
2749 ofproto_reconnect_controllers(br->ofproto);
2751 HMAP_FOR_EACH (br, node, &all_bridges) {
2752 ofproto_reconnect_controllers(br->ofproto);
2755 unixctl_command_reply(conn, NULL);
2759 bridge_get_controllers(const struct bridge *br,
2760 struct ovsrec_controller ***controllersp)
2762 struct ovsrec_controller **controllers;
2763 size_t n_controllers;
2765 controllers = br->cfg->controller;
2766 n_controllers = br->cfg->n_controller;
2768 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2774 *controllersp = controllers;
2776 return n_controllers;
2780 bridge_queue_if_cfg(struct bridge *br,
2781 const struct ovsrec_interface *cfg,
2782 const struct ovsrec_port *parent)
2784 struct if_cfg *if_cfg = xmalloc(sizeof *if_cfg);
2787 if_cfg->parent = parent;
2788 if_cfg->ofport = iface_pick_ofport(cfg);
2789 hmap_insert(&br->if_cfg_todo, &if_cfg->hmap_node,
2790 hash_string(if_cfg->cfg->name, 0));
2793 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2794 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2795 * 'br' needs to complete its configuration. */
2797 bridge_add_del_ports(struct bridge *br,
2798 const unsigned long int *splinter_vlans)
2800 struct shash_node *port_node;
2801 struct port *port, *next;
2802 struct shash new_ports;
2805 ovs_assert(hmap_is_empty(&br->if_cfg_todo));
2807 /* Collect new ports. */
2808 shash_init(&new_ports);
2809 for (i = 0; i < br->cfg->n_ports; i++) {
2810 const char *name = br->cfg->ports[i]->name;
2811 if (!shash_add_once(&new_ports, name, br->cfg->ports[i])) {
2812 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2816 if (bridge_get_controllers(br, NULL)
2817 && !shash_find(&new_ports, br->name)) {
2818 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2819 br->name, br->name);
2821 ovsrec_interface_init(&br->synth_local_iface);
2822 ovsrec_port_init(&br->synth_local_port);
2824 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2825 br->synth_local_port.n_interfaces = 1;
2826 br->synth_local_port.name = br->name;
2828 br->synth_local_iface.name = br->name;
2829 br->synth_local_iface.type = "internal";
2831 br->synth_local_ifacep = &br->synth_local_iface;
2833 shash_add(&new_ports, br->name, &br->synth_local_port);
2836 if (splinter_vlans) {
2837 add_vlan_splinter_ports(br, splinter_vlans, &new_ports);
2840 /* Get rid of deleted ports.
2841 * Get rid of deleted interfaces on ports that still exist. */
2842 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2843 port->cfg = shash_find_data(&new_ports, port->name);
2847 port_del_ifaces(port);
2851 /* Update iface->cfg and iface->type in interfaces that still exist.
2852 * Add new interfaces to creation queue. */
2853 SHASH_FOR_EACH (port_node, &new_ports) {
2854 const struct ovsrec_port *port = port_node->data;
2857 for (i = 0; i < port->n_interfaces; i++) {
2858 const struct ovsrec_interface *cfg = port->interfaces[i];
2859 struct iface *iface = iface_lookup(br, cfg->name);
2860 const char *type = iface_get_type(cfg, br->cfg);
2865 } else if (!strcmp(type, "null")) {
2866 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2867 " may be removed in February 2013. Please email"
2868 " dev@openvswitch.org with concerns.",
2871 bridge_queue_if_cfg(br, cfg, port);
2876 shash_destroy(&new_ports);
2879 /* Initializes 'oc' appropriately as a management service controller for
2882 * The caller must free oc->target when it is no longer needed. */
2884 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2885 struct ofproto_controller *oc)
2887 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2888 oc->max_backoff = 0;
2889 oc->probe_interval = 60;
2890 oc->band = OFPROTO_OUT_OF_BAND;
2892 oc->burst_limit = 0;
2893 oc->enable_async_msgs = true;
2896 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2898 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2899 struct ofproto_controller *oc)
2903 oc->target = c->target;
2904 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2905 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2906 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2907 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2908 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2909 oc->burst_limit = (c->controller_burst_limit
2910 ? *c->controller_burst_limit : 0);
2911 oc->enable_async_msgs = (!c->enable_async_messages
2912 || *c->enable_async_messages);
2913 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2914 if (dscp < 0 || dscp > 63) {
2915 dscp = DSCP_DEFAULT;
2920 /* Configures the IP stack for 'br''s local interface properly according to the
2921 * configuration in 'c'. */
2923 bridge_configure_local_iface_netdev(struct bridge *br,
2924 struct ovsrec_controller *c)
2926 struct netdev *netdev;
2927 struct in_addr mask, gateway;
2929 struct iface *local_iface;
2932 /* If there's no local interface or no IP address, give up. */
2933 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2934 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2938 /* Bring up the local interface. */
2939 netdev = local_iface->netdev;
2940 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2942 /* Configure the IP address and netmask. */
2943 if (!c->local_netmask
2944 || !inet_aton(c->local_netmask, &mask)
2946 mask.s_addr = guess_netmask(ip.s_addr);
2948 if (!netdev_set_in4(netdev, ip, mask)) {
2949 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2950 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2953 /* Configure the default gateway. */
2954 if (c->local_gateway
2955 && inet_aton(c->local_gateway, &gateway)
2956 && gateway.s_addr) {
2957 if (!netdev_add_router(netdev, gateway)) {
2958 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2959 br->name, IP_ARGS(gateway.s_addr));
2964 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2965 * in either string are treated as equal to any number of slashes in the other,
2966 * e.g. "x///y" is equal to "x/y".
2968 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2969 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2970 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2971 * 'b' against a prefix of 'a'.
2974 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2976 const char *b_start = b;
2978 if (b - b_start >= b_stoplen) {
2980 } else if (*a != *b) {
2982 } else if (*a == '/') {
2983 a += strspn(a, "/");
2984 b += strspn(b, "/");
2985 } else if (*a == '\0') {
2995 bridge_configure_remotes(struct bridge *br,
2996 const struct sockaddr_in *managers, size_t n_managers)
2998 bool disable_in_band;
3000 struct ovsrec_controller **controllers;
3001 size_t n_controllers;
3003 enum ofproto_fail_mode fail_mode;
3005 struct ofproto_controller *ocs;
3009 /* Check if we should disable in-band control on this bridge. */
3010 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3013 /* Set OpenFlow queue ID for in-band control. */
3014 ofproto_set_in_band_queue(br->ofproto,
3015 smap_get_int(&br->cfg->other_config,
3016 "in-band-queue", -1));
3018 if (disable_in_band) {
3019 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3021 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3024 n_controllers = bridge_get_controllers(br, &controllers);
3026 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3029 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3030 for (i = 0; i < n_controllers; i++) {
3031 struct ovsrec_controller *c = controllers[i];
3033 if (!strncmp(c->target, "punix:", 6)
3034 || !strncmp(c->target, "unix:", 5)) {
3035 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3038 if (!strncmp(c->target, "unix:", 5)) {
3039 /* Connect to a listening socket */
3040 whitelist = xasprintf("unix:%s/", ovs_rundir());
3041 if (strchr(c->target, '/') &&
3042 !equal_pathnames(c->target, whitelist,
3043 strlen(whitelist))) {
3044 /* Absolute path specified, but not in ovs_rundir */
3045 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3046 "controller \"%s\" due to possibility for "
3047 "remote exploit. Instead, specify socket "
3048 "in whitelisted \"%s\" or connect to "
3049 "\"unix:%s/%s.mgmt\" (which is always "
3050 "available without special configuration).",
3051 br->name, c->target, whitelist,
3052 ovs_rundir(), br->name);
3057 whitelist = xasprintf("punix:%s/%s.controller",
3058 ovs_rundir(), br->name);
3059 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3060 /* Prevent remote ovsdb-server users from accessing
3061 * arbitrary Unix domain sockets and overwriting arbitrary
3063 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3064 "controller \"%s\" due to possibility of "
3065 "overwriting local files. Instead, specify "
3066 "whitelisted \"%s\" or connect to "
3067 "\"unix:%s/%s.mgmt\" (which is always "
3068 "available without special configuration).",
3069 br->name, c->target, whitelist,
3070 ovs_rundir(), br->name);
3079 bridge_configure_local_iface_netdev(br, c);
3080 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3081 if (disable_in_band) {
3082 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3087 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3088 bridge_get_allowed_versions(br));
3089 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3092 /* Set the fail-mode. */
3093 fail_mode = !br->cfg->fail_mode
3094 || !strcmp(br->cfg->fail_mode, "standalone")
3095 ? OFPROTO_FAIL_STANDALONE
3096 : OFPROTO_FAIL_SECURE;
3097 ofproto_set_fail_mode(br->ofproto, fail_mode);
3099 /* Configure OpenFlow controller connection snooping. */
3100 if (!ofproto_has_snoops(br->ofproto)) {
3104 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3105 ovs_rundir(), br->name));
3106 ofproto_set_snoops(br->ofproto, &snoops);
3107 sset_destroy(&snoops);
3112 bridge_configure_tables(struct bridge *br)
3114 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3118 n_tables = ofproto_get_n_tables(br->ofproto);
3120 for (i = 0; i < n_tables; i++) {
3121 struct ofproto_table_settings s;
3124 s.max_flows = UINT_MAX;
3128 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3129 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3132 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3133 s.max_flows = *cfg->flow_limit;
3135 if (cfg->overflow_policy
3136 && !strcmp(cfg->overflow_policy, "evict")) {
3139 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3140 for (k = 0; k < cfg->n_groups; k++) {
3141 const char *string = cfg->groups[k];
3144 msg = mf_parse_subfield__(&s.groups[k], &string);
3146 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3147 "'groups' (%s)", br->name, i, msg);
3149 } else if (*string) {
3150 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3151 "element '%s' contains trailing garbage",
3152 br->name, i, cfg->groups[k]);
3160 ofproto_configure_table(br->ofproto, i, &s);
3164 for (; j < br->cfg->n_flow_tables; j++) {
3165 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3166 "%"PRId64" not supported by this datapath", br->name,
3167 br->cfg->key_flow_tables[j]);
3172 bridge_configure_dp_desc(struct bridge *br)
3174 ofproto_set_dp_desc(br->ofproto,
3175 smap_get(&br->cfg->other_config, "dp-desc"));
3178 /* Port functions. */
3180 static struct port *
3181 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3185 port = xzalloc(sizeof *port);
3187 port->name = xstrdup(cfg->name);
3189 list_init(&port->ifaces);
3191 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3195 /* Deletes interfaces from 'port' that are no longer configured for it. */
3197 port_del_ifaces(struct port *port)
3199 struct iface *iface, *next;
3200 struct sset new_ifaces;
3203 /* Collect list of new interfaces. */
3204 sset_init(&new_ifaces);
3205 for (i = 0; i < port->cfg->n_interfaces; i++) {
3206 const char *name = port->cfg->interfaces[i]->name;
3207 const char *type = port->cfg->interfaces[i]->type;
3208 if (strcmp(type, "null")) {
3209 sset_add(&new_ifaces, name);
3213 /* Get rid of deleted interfaces. */
3214 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3215 if (!sset_contains(&new_ifaces, iface->name)) {
3216 iface_destroy(iface);
3220 sset_destroy(&new_ifaces);
3224 port_destroy(struct port *port)
3227 struct bridge *br = port->bridge;
3228 struct iface *iface, *next;
3231 ofproto_bundle_unregister(br->ofproto, port);
3234 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3235 iface_destroy(iface);
3238 hmap_remove(&br->ports, &port->hmap_node);
3244 static struct port *
3245 port_lookup(const struct bridge *br, const char *name)
3249 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3251 if (!strcmp(port->name, name)) {
3259 enable_lacp(struct port *port, bool *activep)
3261 if (!port->cfg->lacp) {
3262 /* XXX when LACP implementation has been sufficiently tested, enable by
3263 * default and make active on bonded ports. */
3265 } else if (!strcmp(port->cfg->lacp, "off")) {
3267 } else if (!strcmp(port->cfg->lacp, "active")) {
3270 } else if (!strcmp(port->cfg->lacp, "passive")) {
3274 VLOG_WARN("port %s: unknown LACP mode %s",
3275 port->name, port->cfg->lacp);
3280 static struct lacp_settings *
3281 port_configure_lacp(struct port *port, struct lacp_settings *s)
3283 const char *lacp_time, *system_id;
3286 if (!enable_lacp(port, &s->active)) {
3290 s->name = port->name;
3292 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3294 if (sscanf(system_id, ETH_ADDR_SCAN_FMT,
3295 ETH_ADDR_SCAN_ARGS(s->id)) != ETH_ADDR_SCAN_COUNT) {
3296 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3297 " address.", port->name, system_id);
3301 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3304 if (eth_addr_is_zero(s->id)) {
3305 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3309 /* Prefer bondable links if unspecified. */
3310 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3312 s->priority = (priority > 0 && priority <= UINT16_MAX
3314 : UINT16_MAX - !list_is_short(&port->ifaces));
3316 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3317 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3322 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3324 int priority, portid, key;
3326 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3327 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3329 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3331 if (portid <= 0 || portid > UINT16_MAX) {
3332 portid = ofp_to_u16(iface->ofp_port);
3335 if (priority <= 0 || priority > UINT16_MAX) {
3336 priority = UINT16_MAX;
3339 if (key < 0 || key > UINT16_MAX) {
3343 s->name = iface->name;
3345 s->priority = priority;
3350 port_configure_bond(struct port *port, struct bond_settings *s)
3352 const char *detect_s;
3353 struct iface *iface;
3354 int miimon_interval;
3356 s->name = port->name;
3358 if (port->cfg->bond_mode) {
3359 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3360 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3361 port->name, port->cfg->bond_mode,
3362 bond_mode_to_string(s->balance));
3365 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3367 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3368 * active-backup. At some point we should remove this warning. */
3369 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3370 " in previous versions, the default bond_mode was"
3371 " balance-slb", port->name,
3372 bond_mode_to_string(s->balance));
3374 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3375 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3376 "please use another bond type or disable flood_vlans",
3380 miimon_interval = smap_get_int(&port->cfg->other_config,
3381 "bond-miimon-interval", 0);
3382 if (miimon_interval <= 0) {
3383 miimon_interval = 200;
3386 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3387 if (!detect_s || !strcmp(detect_s, "carrier")) {
3388 miimon_interval = 0;
3389 } else if (strcmp(detect_s, "miimon")) {
3390 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3391 "defaulting to carrier", port->name, detect_s);
3392 miimon_interval = 0;
3395 s->up_delay = MAX(0, port->cfg->bond_updelay);
3396 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3397 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3398 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3399 "bond-rebalance-interval", 10000);
3400 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3401 s->rebalance_interval = 1000;
3404 s->fake_iface = port->cfg->bond_fake_iface;
3406 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3407 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3411 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3412 * instead of obtaining it from the database. */
3414 port_is_synthetic(const struct port *port)
3416 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3419 /* Interface functions. */
3422 iface_is_internal(const struct ovsrec_interface *iface,
3423 const struct ovsrec_bridge *br)
3425 /* The local port and "internal" ports are always "internal". */
3426 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3429 /* Returns the correct network device type for interface 'iface' in bridge
3432 iface_get_type(const struct ovsrec_interface *iface,
3433 const struct ovsrec_bridge *br)
3437 /* The local port always has type "internal". Other ports take
3438 * their type from the database and default to "system" if none is
3440 if (iface_is_internal(iface, br)) {
3443 type = iface->type[0] ? iface->type : "system";
3446 return ofproto_port_open_type(br->datapath_type, type);
3450 iface_destroy(struct iface *iface)
3453 struct port *port = iface->port;
3454 struct bridge *br = port->bridge;
3456 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3457 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3460 if (iface->ofp_port != OFPP_NONE) {
3461 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3464 list_remove(&iface->port_elem);
3465 hmap_remove(&br->iface_by_name, &iface->name_node);
3467 netdev_close(iface->netdev);
3474 static struct iface *
3475 iface_lookup(const struct bridge *br, const char *name)
3477 struct iface *iface;
3479 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3480 &br->iface_by_name) {
3481 if (!strcmp(iface->name, name)) {
3489 static struct iface *
3490 iface_find(const char *name)
3492 const struct bridge *br;
3494 HMAP_FOR_EACH (br, node, &all_bridges) {
3495 struct iface *iface = iface_lookup(br, name);
3504 static struct if_cfg *
3505 if_cfg_lookup(const struct bridge *br, const char *name)
3507 struct if_cfg *if_cfg;
3509 HMAP_FOR_EACH_WITH_HASH (if_cfg, hmap_node, hash_string(name, 0),
3511 if (!strcmp(if_cfg->cfg->name, name)) {
3519 static struct iface *
3520 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3522 struct iface *iface;
3524 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3526 if (iface->ofp_port == ofp_port) {
3533 /* Set Ethernet address of 'iface', if one is specified in the configuration
3536 iface_set_mac(struct iface *iface)
3538 uint8_t ea[ETH_ADDR_LEN];
3540 if (!strcmp(iface->type, "internal")
3541 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3542 if (iface->ofp_port == OFPP_LOCAL) {
3543 VLOG_ERR("interface %s: ignoring mac in Interface record "
3544 "(use Bridge record to set local port's mac)",
3546 } else if (eth_addr_is_multicast(ea)) {
3547 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3550 int error = netdev_set_etheraddr(iface->netdev, ea);
3552 VLOG_ERR("interface %s: setting MAC failed (%s)",
3553 iface->name, ovs_strerror(error));
3559 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3561 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3564 port_ = (ofport == OFPP_NONE) ? -1 : ofp_to_u16(ofport);
3565 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3566 ovsrec_interface_set_ofport(if_cfg, &port_, 1);
3570 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3571 * sets the "ofport" field to -1.
3573 * This is appropriate when 'if_cfg''s interface cannot be created or is
3574 * otherwise invalid. */
3576 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3578 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3579 ovsrec_interface_set_status(if_cfg, NULL);
3580 ovsrec_interface_set_admin_state(if_cfg, NULL);
3581 ovsrec_interface_set_duplex(if_cfg, NULL);
3582 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3583 ovsrec_interface_set_link_state(if_cfg, NULL);
3584 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3585 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3586 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3587 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3588 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3589 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3590 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3591 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3595 struct iface_delete_queues_cbdata {
3596 struct netdev *netdev;
3597 const struct ovsdb_datum *queues;
3601 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3603 union ovsdb_atom atom;
3605 atom.integer = target;
3606 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3610 iface_delete_queues(unsigned int queue_id,
3611 const struct smap *details OVS_UNUSED, void *cbdata_)
3613 struct iface_delete_queues_cbdata *cbdata = cbdata_;
3615 if (!queue_ids_include(cbdata->queues, queue_id)) {
3616 netdev_delete_queue(cbdata->netdev, queue_id);
3621 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3623 struct ofpbuf queues_buf;
3625 ofpbuf_init(&queues_buf, 0);
3627 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3628 netdev_set_qos(iface->netdev, NULL, NULL);
3630 struct iface_delete_queues_cbdata cbdata;
3634 /* Configure top-level Qos for 'iface'. */
3635 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3637 /* Deconfigure queues that were deleted. */
3638 cbdata.netdev = iface->netdev;
3639 cbdata.queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3641 netdev_dump_queues(iface->netdev, iface_delete_queues, &cbdata);
3643 /* Configure queues for 'iface'. */
3645 for (i = 0; i < qos->n_queues; i++) {
3646 const struct ovsrec_queue *queue = qos->value_queues[i];
3647 unsigned int queue_id = qos->key_queues[i];
3649 if (queue_id == 0) {
3653 if (queue->n_dscp == 1) {
3654 struct ofproto_port_queue *port_queue;
3656 port_queue = ofpbuf_put_uninit(&queues_buf,
3657 sizeof *port_queue);
3658 port_queue->queue = queue_id;
3659 port_queue->dscp = queue->dscp[0];
3662 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3665 struct smap details;
3667 smap_init(&details);
3668 netdev_set_queue(iface->netdev, 0, &details);
3669 smap_destroy(&details);
3673 if (iface->ofp_port != OFPP_NONE) {
3674 const struct ofproto_port_queue *port_queues = queues_buf.data;
3675 size_t n_queues = queues_buf.size / sizeof *port_queues;
3677 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3678 port_queues, n_queues);
3681 netdev_set_policing(iface->netdev,
3682 iface->cfg->ingress_policing_rate,
3683 iface->cfg->ingress_policing_burst);
3685 ofpbuf_uninit(&queues_buf);
3689 iface_configure_cfm(struct iface *iface)
3691 const struct ovsrec_interface *cfg = iface->cfg;
3692 const char *opstate_str;
3693 const char *cfm_ccm_vlan;
3694 struct cfm_settings s;
3695 struct smap netdev_args;
3697 if (!cfg->n_cfm_mpid) {
3698 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3702 s.check_tnl_key = false;
3703 smap_init(&netdev_args);
3704 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3705 const char *key = smap_get(&netdev_args, "key");
3706 const char *in_key = smap_get(&netdev_args, "in_key");
3708 s.check_tnl_key = (key && !strcmp(key, "flow"))
3709 || (in_key && !strcmp(in_key, "flow"));
3711 smap_destroy(&netdev_args);
3713 s.mpid = *cfg->cfm_mpid;
3714 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3715 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3716 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3718 if (s.interval <= 0) {
3722 if (!cfm_ccm_vlan) {
3724 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3725 s.ccm_vlan = CFM_RANDOM_VLAN;
3727 s.ccm_vlan = atoi(cfm_ccm_vlan);
3728 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3733 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3735 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3737 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3738 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3740 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3743 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3744 * instead of obtaining it from the database. */
3746 iface_is_synthetic(const struct iface *iface)
3748 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3752 iface_pick_ofport(const struct ovsrec_interface *cfg)
3754 ofp_port_t ofport = cfg->n_ofport ? u16_to_ofp(*cfg->ofport)
3756 return cfg->n_ofport_request ? u16_to_ofp(*cfg->ofport_request)
3761 /* Port mirroring. */
3763 static struct mirror *
3764 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3768 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3769 if (uuid_equals(uuid, &m->uuid)) {
3777 bridge_configure_mirrors(struct bridge *br)
3779 const struct ovsdb_datum *mc;
3780 unsigned long *flood_vlans;
3781 struct mirror *m, *next;
3784 /* Get rid of deleted mirrors. */
3785 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3786 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3787 union ovsdb_atom atom;
3789 atom.uuid = m->uuid;
3790 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3795 /* Add new mirrors and reconfigure existing ones. */
3796 for (i = 0; i < br->cfg->n_mirrors; i++) {
3797 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3798 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3800 m = mirror_create(br, cfg);
3803 if (!mirror_configure(m)) {
3808 /* Update flooded vlans (for RSPAN). */
3809 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3810 br->cfg->n_flood_vlans);
3811 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3812 bitmap_free(flood_vlans);
3815 static struct mirror *
3816 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3820 m = xzalloc(sizeof *m);
3821 m->uuid = cfg->header_.uuid;
3822 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3824 m->name = xstrdup(cfg->name);
3830 mirror_destroy(struct mirror *m)
3833 struct bridge *br = m->bridge;
3836 ofproto_mirror_unregister(br->ofproto, m);
3839 hmap_remove(&br->mirrors, &m->hmap_node);
3846 mirror_collect_ports(struct mirror *m,
3847 struct ovsrec_port **in_ports, int n_in_ports,
3848 void ***out_portsp, size_t *n_out_portsp)
3850 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3851 size_t n_out_ports = 0;
3854 for (i = 0; i < n_in_ports; i++) {
3855 const char *name = in_ports[i]->name;
3856 struct port *port = port_lookup(m->bridge, name);
3858 out_ports[n_out_ports++] = port;
3860 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3861 "port %s", m->bridge->name, m->name, name);
3864 *out_portsp = out_ports;
3865 *n_out_portsp = n_out_ports;
3869 mirror_configure(struct mirror *m)
3871 const struct ovsrec_mirror *cfg = m->cfg;
3872 struct ofproto_mirror_settings s;
3875 if (strcmp(cfg->name, m->name)) {
3877 m->name = xstrdup(cfg->name);
3881 /* Get output port or VLAN. */
3882 if (cfg->output_port) {
3883 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3884 if (!s.out_bundle) {
3885 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3886 m->bridge->name, m->name);
3889 s.out_vlan = UINT16_MAX;
3891 if (cfg->output_vlan) {
3892 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3893 "output vlan; ignoring output vlan",
3894 m->bridge->name, m->name);
3896 } else if (cfg->output_vlan) {
3897 /* The database should prevent invalid VLAN values. */
3898 s.out_bundle = NULL;
3899 s.out_vlan = *cfg->output_vlan;
3901 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3902 m->bridge->name, m->name);
3906 /* Get port selection. */
3907 if (cfg->select_all) {
3908 size_t n_ports = hmap_count(&m->bridge->ports);
3909 void **ports = xmalloc(n_ports * sizeof *ports);
3914 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3924 /* Get ports, dropping ports that don't exist.
3925 * The IDL ensures that there are no duplicates. */
3926 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3927 &s.srcs, &s.n_srcs);
3928 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3929 &s.dsts, &s.n_dsts);
3932 /* Get VLAN selection. */
3933 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3936 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3939 if (s.srcs != s.dsts) {
3948 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3950 * This is deprecated. It is only for compatibility with broken device drivers
3951 * in old versions of Linux that do not properly support VLANs when VLAN
3952 * devices are not used. When broken device drivers are no longer in
3953 * widespread use, we will delete these interfaces. */
3955 static struct ovsrec_port **recs;
3956 static size_t n_recs, allocated_recs;
3958 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3959 * splinters are reconfigured. */
3961 register_rec(struct ovsrec_port *rec)
3963 if (n_recs >= allocated_recs) {
3964 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3966 recs[n_recs++] = rec;
3969 /* Frees all of the ports registered with register_reg(). */
3971 free_registered_recs(void)
3975 for (i = 0; i < n_recs; i++) {
3976 struct ovsrec_port *port = recs[i];
3979 for (j = 0; j < port->n_interfaces; j++) {
3980 struct ovsrec_interface *iface = port->interfaces[j];
3985 smap_destroy(&port->other_config);
3986 free(port->interfaces);
3994 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3997 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
3999 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4003 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4006 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4007 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4008 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4011 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4012 * use, returns NULL.
4014 * Updates 'vlan_splinters_enabled_anywhere'. */
4015 static unsigned long int *
4016 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4018 unsigned long int *splinter_vlans;
4019 struct sset splinter_ifaces;
4020 const char *real_dev_name;
4021 struct shash *real_devs;
4022 struct shash_node *node;
4026 /* Free space allocated for synthesized ports and interfaces, since we're
4027 * in the process of reconstructing all of them. */
4028 free_registered_recs();
4030 splinter_vlans = bitmap_allocate(4096);
4031 sset_init(&splinter_ifaces);
4032 vlan_splinters_enabled_anywhere = false;
4033 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4034 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4037 for (j = 0; j < br_cfg->n_ports; j++) {
4038 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4041 for (k = 0; k < port_cfg->n_interfaces; k++) {
4042 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4044 if (vlan_splinters_is_enabled(iface_cfg)) {
4045 vlan_splinters_enabled_anywhere = true;
4046 sset_add(&splinter_ifaces, iface_cfg->name);
4047 vlan_bitmap_from_array__(port_cfg->trunks,
4053 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4054 bitmap_set1(splinter_vlans, *port_cfg->tag);
4059 if (!vlan_splinters_enabled_anywhere) {
4060 free(splinter_vlans);
4061 sset_destroy(&splinter_ifaces);
4065 HMAP_FOR_EACH (br, node, &all_bridges) {
4067 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4071 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4072 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4073 * device to be created for it. */
4074 bitmap_set0(splinter_vlans, 0);
4075 bitmap_set0(splinter_vlans, 4095);
4077 /* Delete all VLAN devices that we don't need. */
4079 real_devs = vlandev_get_real_devs();
4080 SHASH_FOR_EACH (node, real_devs) {
4081 const struct vlan_real_dev *real_dev = node->data;
4082 const struct vlan_dev *vlan_dev;
4083 bool real_dev_has_splinters;
4085 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4087 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4088 if (!real_dev_has_splinters
4089 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4090 struct netdev *netdev;
4092 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4093 if (!netdev_get_in4(netdev, NULL, NULL) ||
4094 !netdev_get_in6(netdev, NULL)) {
4095 /* It has an IP address configured, so we don't own
4096 * it. Don't delete it. */
4098 vlandev_del(vlan_dev->name);
4100 netdev_close(netdev);
4107 /* Add all VLAN devices that we need. */
4108 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4111 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4112 if (!vlandev_get_name(real_dev_name, vid)) {
4113 vlandev_add(real_dev_name, vid);
4120 sset_destroy(&splinter_ifaces);
4122 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4123 free(splinter_vlans);
4126 return splinter_vlans;
4129 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4132 configure_splinter_port(struct port *port)
4134 struct ofproto *ofproto = port->bridge->ofproto;
4135 ofp_port_t realdev_ofp_port;
4136 const char *realdev_name;
4137 struct iface *vlandev, *realdev;
4139 ofproto_bundle_unregister(port->bridge->ofproto, port);
4141 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4144 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4145 realdev = iface_lookup(port->bridge, realdev_name);
4146 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4148 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4152 static struct ovsrec_port *
4153 synthesize_splinter_port(const char *real_dev_name,
4154 const char *vlan_dev_name, int vid)
4156 struct ovsrec_interface *iface;
4157 struct ovsrec_port *port;
4159 iface = xmalloc(sizeof *iface);
4160 ovsrec_interface_init(iface);
4161 iface->name = xstrdup(vlan_dev_name);
4162 iface->type = "system";
4164 port = xmalloc(sizeof *port);
4165 ovsrec_port_init(port);
4166 port->interfaces = xmemdup(&iface, sizeof iface);
4167 port->n_interfaces = 1;
4168 port->name = xstrdup(vlan_dev_name);
4169 port->vlan_mode = "splinter";
4170 port->tag = xmalloc(sizeof *port->tag);
4173 smap_add(&port->other_config, "realdev", real_dev_name);
4179 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4180 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4181 * 1-bit in the 'splinter_vlans' bitmap. */
4183 add_vlan_splinter_ports(struct bridge *br,
4184 const unsigned long int *splinter_vlans,
4185 struct shash *ports)
4189 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4190 * we're modifying 'ports'. */
4191 for (i = 0; i < br->cfg->n_ports; i++) {
4192 const char *name = br->cfg->ports[i]->name;
4193 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4196 for (j = 0; j < port_cfg->n_interfaces; j++) {
4197 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4199 if (vlan_splinters_is_enabled(iface_cfg)) {
4200 const char *real_dev_name;
4203 real_dev_name = iface_cfg->name;
4204 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4205 const char *vlan_dev_name;
4207 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4209 && !shash_find(ports, vlan_dev_name)) {
4210 shash_add(ports, vlan_dev_name,
4211 synthesize_splinter_port(
4212 real_dev_name, vlan_dev_name, vid));
4221 mirror_refresh_stats(struct mirror *m)
4223 struct ofproto *ofproto = m->bridge->ofproto;
4224 uint64_t tx_packets, tx_bytes;
4227 size_t stat_cnt = 0;
4229 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4230 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4234 if (tx_packets != UINT64_MAX) {
4235 keys[stat_cnt] = "tx_packets";
4236 values[stat_cnt] = tx_packets;
4239 if (tx_bytes != UINT64_MAX) {
4240 keys[stat_cnt] = "tx_bytes";
4241 values[stat_cnt] = tx_bytes;
4245 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);