1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 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"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge);
66 COVERAGE_DEFINE(bridge_reconfigure);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
82 /* These members are valid only within bridge_reconfigure(). */
83 const char *type; /* Usually same as cfg->type. */
84 const struct ovsrec_interface *cfg;
88 struct uuid uuid; /* UUID of this "mirror" record in database. */
89 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
90 struct bridge *bridge;
92 const struct ovsrec_mirror *cfg;
96 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
97 struct bridge *bridge;
100 const struct ovsrec_port *cfg;
102 /* An ordinary bridge port has 1 interface.
103 * A bridge port for bonding has at least 2 interfaces. */
104 struct list ifaces; /* List of "struct iface"s. */
108 struct hmap_node node; /* In 'all_bridges'. */
109 char *name; /* User-specified arbitrary name. */
110 char *type; /* Datapath type. */
111 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
112 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
113 const struct ovsrec_bridge *cfg;
115 /* OpenFlow switch processing. */
116 struct ofproto *ofproto; /* OpenFlow switch. */
119 struct hmap ports; /* "struct port"s indexed by name. */
120 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
121 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
123 /* Port mirroring. */
124 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
126 /* Used during reconfiguration. */
127 struct shash wanted_ports;
129 /* Synthetic local port if necessary. */
130 struct ovsrec_port synth_local_port;
131 struct ovsrec_interface synth_local_iface;
132 struct ovsrec_interface *synth_local_ifacep;
135 /* All bridges, indexed by name. */
136 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
138 /* OVSDB IDL used to obtain configuration. */
139 static struct ovsdb_idl *idl;
141 /* We want to complete daemonization, fully detaching from our parent process,
142 * only after we have completed our initial configuration, committed our state
143 * to the database, and received confirmation back from the database server
144 * that it applied the commit. This allows our parent process to know that,
145 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
146 * to have already been filled in. (It is only "very likely" rather than
147 * certain because there is always a slim possibility that the transaction will
148 * fail or that some other client has added new bridges, ports, etc. while
149 * ovs-vswitchd was configuring using an old configuration.)
151 * We only need to do this once for our initial configuration at startup, so
152 * 'initial_config_done' tracks whether we've already done it. While we are
153 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
154 * itself and is otherwise NULL. */
155 static bool initial_config_done;
156 static struct ovsdb_idl_txn *daemonize_txn;
158 /* Most recently processed IDL sequence number. */
159 static unsigned int idl_seqno;
161 /* Track changes to port connectivity. */
162 static uint64_t connectivity_seqno = LLONG_MIN;
164 /* Status update to database.
166 * Some information in the database must be kept as up-to-date as possible to
167 * allow controllers to respond rapidly to network outages. Those status are
168 * updated via the 'status_txn'.
170 * We use the global connectivity sequence number to detect the status change.
171 * Also, to prevent the status update from sending too much to the database,
172 * we check the return status of each update transaction and do not start new
173 * update if the previous transaction status is 'TXN_INCOMPLETE'.
175 * 'statux_txn' is NULL if there is no ongoing status update.
177 static struct ovsdb_idl_txn *status_txn;
179 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
180 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
181 #define STATUS_CHECK_AGAIN_MSEC 100
183 /* Each time this timer expires, the bridge fetches interface and mirror
184 * statistics and pushes them into the database. */
185 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
186 static long long int iface_stats_timer = LLONG_MIN;
188 /* Set to true to allow experimental use of OpenFlow 1.4.
189 * This is false initially because OpenFlow 1.4 is not yet safe to use: it can
190 * abort due to unimplemented features. */
191 static bool allow_of14;
193 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
194 * expensive. This can cause bridge_reconfigure() to take a long time during
195 * which no other work can be done. To deal with this problem, we limit port
196 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
197 * call to bridge_reconfigure(). If there is more work to do after the limit
198 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
199 * This allows the rest of the code to catch up on important things like
200 * forwarding packets. */
201 #define OFP_PORT_ACTION_WINDOW 10
203 static void add_del_bridges(const struct ovsrec_open_vswitch *);
204 static void bridge_run__(void);
205 static void bridge_create(const struct ovsrec_bridge *);
206 static void bridge_destroy(struct bridge *);
207 static struct bridge *bridge_lookup(const char *name);
208 static unixctl_cb_func bridge_unixctl_dump_flows;
209 static unixctl_cb_func bridge_unixctl_reconnect;
210 static size_t bridge_get_controllers(const struct bridge *br,
211 struct ovsrec_controller ***controllersp);
212 static void bridge_collect_wanted_ports(struct bridge *,
213 const unsigned long *splinter_vlans,
214 struct shash *wanted_ports);
215 static void bridge_delete_ofprotos(void);
216 static void bridge_delete_or_reconfigure_ports(struct bridge *);
217 static void bridge_del_ports(struct bridge *,
218 const struct shash *wanted_ports);
219 static void bridge_add_ports(struct bridge *,
220 const struct shash *wanted_ports);
222 static void bridge_configure_datapath_id(struct bridge *);
223 static void bridge_configure_netflow(struct bridge *);
224 static void bridge_configure_forward_bpdu(struct bridge *);
225 static void bridge_configure_mac_table(struct bridge *);
226 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
227 static void bridge_configure_ipfix(struct bridge *);
228 static void bridge_configure_stp(struct bridge *);
229 static void bridge_configure_tables(struct bridge *);
230 static void bridge_configure_dp_desc(struct bridge *);
231 static void bridge_configure_remotes(struct bridge *,
232 const struct sockaddr_in *managers,
234 static void bridge_pick_local_hw_addr(struct bridge *,
235 uint8_t ea[ETH_ADDR_LEN],
236 struct iface **hw_addr_iface);
237 static uint64_t bridge_pick_datapath_id(struct bridge *,
238 const uint8_t bridge_ea[ETH_ADDR_LEN],
239 struct iface *hw_addr_iface);
240 static uint64_t dpid_from_hash(const void *, size_t nbytes);
241 static bool bridge_has_bond_fake_iface(const struct bridge *,
243 static bool port_is_bond_fake_iface(const struct port *);
245 static unixctl_cb_func qos_unixctl_show;
247 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
248 static void port_del_ifaces(struct port *);
249 static void port_destroy(struct port *);
250 static struct port *port_lookup(const struct bridge *, const char *name);
251 static void port_configure(struct port *);
252 static struct lacp_settings *port_configure_lacp(struct port *,
253 struct lacp_settings *);
254 static void port_configure_bond(struct port *, struct bond_settings *);
255 static bool port_is_synthetic(const struct port *);
257 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
258 static void run_system_stats(void);
260 static void bridge_configure_mirrors(struct bridge *);
261 static struct mirror *mirror_create(struct bridge *,
262 const struct ovsrec_mirror *);
263 static void mirror_destroy(struct mirror *);
264 static bool mirror_configure(struct mirror *);
265 static void mirror_refresh_stats(struct mirror *);
267 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
268 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
269 const struct ovsrec_port *);
270 static bool iface_is_internal(const struct ovsrec_interface *iface,
271 const struct ovsrec_bridge *br);
272 static const char *iface_get_type(const struct ovsrec_interface *,
273 const struct ovsrec_bridge *);
274 static void iface_destroy(struct iface *);
275 static void iface_destroy__(struct iface *);
276 static struct iface *iface_lookup(const struct bridge *, const char *name);
277 static struct iface *iface_find(const char *name);
278 static struct iface *iface_from_ofp_port(const struct bridge *,
279 ofp_port_t ofp_port);
280 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
281 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
282 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
283 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
284 static void iface_configure_cfm(struct iface *);
285 static void iface_refresh_cfm_stats(struct iface *);
286 static void iface_refresh_stats(struct iface *);
287 static void iface_refresh_netdev_status(struct iface *);
288 static void iface_refresh_ofproto_status(struct iface *);
289 static bool iface_is_synthetic(const struct iface *);
290 static ofp_port_t iface_get_requested_ofp_port(
291 const struct ovsrec_interface *);
292 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
294 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
296 * This is deprecated. It is only for compatibility with broken device drivers
297 * in old versions of Linux that do not properly support VLANs when VLAN
298 * devices are not used. When broken device drivers are no longer in
299 * widespread use, we will delete these interfaces. */
301 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
302 static bool vlan_splinters_enabled_anywhere;
304 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
305 static unsigned long int *collect_splinter_vlans(
306 const struct ovsrec_open_vswitch *);
307 static void configure_splinter_port(struct port *);
308 static void add_vlan_splinter_ports(struct bridge *,
309 const unsigned long int *splinter_vlans,
310 struct shash *ports);
313 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
315 struct shash iface_hints;
316 static bool initialized = false;
323 shash_init(&iface_hints);
326 for (i = 0; i < cfg->n_bridges; i++) {
327 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
330 for (j = 0; j < br_cfg->n_ports; j++) {
331 struct ovsrec_port *port_cfg = br_cfg->ports[j];
334 for (k = 0; k < port_cfg->n_interfaces; k++) {
335 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
336 struct iface_hint *iface_hint;
338 iface_hint = xmalloc(sizeof *iface_hint);
339 iface_hint->br_name = br_cfg->name;
340 iface_hint->br_type = br_cfg->datapath_type;
341 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
343 shash_add(&iface_hints, if_cfg->name, iface_hint);
349 ofproto_init(&iface_hints);
351 shash_destroy_free_data(&iface_hints);
355 /* Public functions. */
357 /* Initializes the bridge module, configuring it to obtain its configuration
358 * from an OVSDB server accessed over 'remote', which should be a string in a
359 * form acceptable to ovsdb_idl_create(). */
361 bridge_init(const char *remote)
363 /* Create connection to database. */
364 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
365 idl_seqno = ovsdb_idl_get_seqno(idl);
366 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
367 ovsdb_idl_verify_write_only(idl);
369 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
370 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
371 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
372 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
373 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
374 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
375 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
377 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
378 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
379 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
381 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
382 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
383 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
386 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
387 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
388 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
389 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
404 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
406 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
407 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
408 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
409 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
411 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
413 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
415 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
416 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
418 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
419 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
420 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
421 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
423 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
425 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
426 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
427 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
429 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
431 /* Register unixctl commands. */
432 unixctl_command_register("qos/show", "interface", 1, 1,
433 qos_unixctl_show, NULL);
434 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
435 bridge_unixctl_dump_flows, NULL);
436 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
437 bridge_unixctl_reconnect, NULL);
447 struct bridge *br, *next_br;
449 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
452 ovsdb_idl_destroy(idl);
455 /* Enables use of OpenFlow 1.4. This is off by default because OpenFlow 1.4 is
456 * not yet safe to use: it can abort due to unimplemented features. */
458 bridge_enable_of14(void)
463 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
464 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
465 * responsible for freeing '*managersp' (with free()).
467 * You may be asking yourself "why does ovs-vswitchd care?", because
468 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
469 * should not be and in fact is not directly involved in that. But
470 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
471 * it has to tell in-band control where the managers are to enable that.
472 * (Thus, only managers connected in-band are collected.)
475 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
476 struct sockaddr_in **managersp, size_t *n_managersp)
478 struct sockaddr_in *managers = NULL;
479 size_t n_managers = 0;
483 /* Collect all of the potential targets from the "targets" columns of the
484 * rows pointed to by "manager_options", excluding any that are
487 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
488 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
490 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
491 sset_find_and_delete(&targets, m->target);
493 sset_add(&targets, m->target);
497 /* Now extract the targets' IP addresses. */
498 if (!sset_is_empty(&targets)) {
501 managers = xmalloc(sset_count(&targets) * sizeof *managers);
502 SSET_FOR_EACH (target, &targets) {
503 struct sockaddr_storage ss;
505 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
507 && ss.ss_family == AF_INET) {
508 managers[n_managers++] = *(struct sockaddr_in *) &ss;
512 sset_destroy(&targets);
514 *managersp = managers;
515 *n_managersp = n_managers;
519 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
521 unsigned long int *splinter_vlans;
522 struct sockaddr_in *managers;
523 struct bridge *br, *next;
524 int sflow_bridge_number;
527 COVERAGE_INC(bridge_reconfigure);
529 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
530 OFPROTO_FLOW_LIMIT_DEFAULT));
531 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
532 OFPROTO_MAX_IDLE_DEFAULT));
535 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
536 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
538 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
539 * to 'ovs_cfg', with only very minimal configuration otherwise.
541 * This is mostly an update to bridge data structures. Nothing is pushed
542 * down to ofproto or lower layers. */
543 add_del_bridges(ovs_cfg);
544 splinter_vlans = collect_splinter_vlans(ovs_cfg);
545 HMAP_FOR_EACH (br, node, &all_bridges) {
546 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
547 bridge_del_ports(br, &br->wanted_ports);
549 free(splinter_vlans);
551 /* Start pushing configuration changes down to the ofproto layer:
553 * - Delete ofprotos that are no longer configured.
555 * - Delete ports that are no longer configured.
557 * - Reconfigure existing ports to their desired configurations, or
558 * delete them if not possible.
560 * We have to do all the deletions before we can do any additions, because
561 * the ports to be added might require resources that will be freed up by
562 * deletions (they might especially overlap in name). */
563 bridge_delete_ofprotos();
564 HMAP_FOR_EACH (br, node, &all_bridges) {
566 bridge_delete_or_reconfigure_ports(br);
570 /* Finish pushing configuration changes to the ofproto layer:
572 * - Create ofprotos that are missing.
574 * - Add ports that are missing. */
575 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
579 error = ofproto_create(br->name, br->type, &br->ofproto);
581 VLOG_ERR("failed to create bridge %s: %s", br->name,
582 ovs_strerror(error));
583 shash_destroy(&br->wanted_ports);
588 HMAP_FOR_EACH (br, node, &all_bridges) {
589 bridge_add_ports(br, &br->wanted_ports);
590 shash_destroy(&br->wanted_ports);
593 reconfigure_system_stats(ovs_cfg);
595 /* Complete the configuration. */
596 sflow_bridge_number = 0;
597 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
598 HMAP_FOR_EACH (br, node, &all_bridges) {
601 /* We need the datapath ID early to allow LACP ports to use it as the
602 * default system ID. */
603 bridge_configure_datapath_id(br);
605 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
608 port_configure(port);
610 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
611 iface_set_ofport(iface->cfg, iface->ofp_port);
612 iface_configure_cfm(iface);
613 iface_configure_qos(iface, port->cfg->qos);
614 iface_set_mac(br, port, iface);
615 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
619 bridge_configure_mirrors(br);
620 bridge_configure_forward_bpdu(br);
621 bridge_configure_mac_table(br);
622 bridge_configure_remotes(br, managers, n_managers);
623 bridge_configure_netflow(br);
624 bridge_configure_sflow(br, &sflow_bridge_number);
625 bridge_configure_ipfix(br);
626 bridge_configure_stp(br);
627 bridge_configure_tables(br);
628 bridge_configure_dp_desc(br);
632 /* The ofproto-dpif provider does some final reconfiguration in its
633 * ->type_run() function. We have to call it before notifying the database
634 * client that reconfiguration is complete, otherwise there is a very
635 * narrow race window in which e.g. ofproto/trace will not recognize the
636 * new configuration (sometimes this causes unit test failures). */
640 /* Delete ofprotos which aren't configured or have the wrong type. Create
641 * ofprotos which don't exist but need to. */
643 bridge_delete_ofprotos(void)
650 /* Delete ofprotos with no bridge or with the wrong type. */
653 ofproto_enumerate_types(&types);
654 SSET_FOR_EACH (type, &types) {
657 ofproto_enumerate_names(type, &names);
658 SSET_FOR_EACH (name, &names) {
659 br = bridge_lookup(name);
660 if (!br || strcmp(type, br->type)) {
661 ofproto_delete(name, type);
665 sset_destroy(&names);
666 sset_destroy(&types);
670 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
672 if (*n >= *allocated) {
673 ports = x2nrealloc(ports, allocated, sizeof *ports);
675 ports[(*n)++] = port;
680 bridge_delete_or_reconfigure_ports(struct bridge *br)
682 struct ofproto_port ofproto_port;
683 struct ofproto_port_dump dump;
685 struct sset ofproto_ports;
686 struct port *port, *port_next;
688 /* List of "ofp_port"s to delete. We make a list instead of deleting them
689 * right away because ofproto implementations aren't necessarily able to
690 * iterate through a changing list of ports in an entirely robust way. */
697 sset_init(&ofproto_ports);
699 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
700 * that are not configured in the database. (This commonly happens when
701 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
703 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
704 * that have the wrong OpenFlow port number (and arrange to add them back
705 * with the correct OpenFlow port number). */
706 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
707 ofp_port_t requested_ofp_port;
710 sset_add(&ofproto_ports, ofproto_port.name);
712 iface = iface_lookup(br, ofproto_port.name);
714 /* No such iface is configured, so we should delete this
717 * As a corner case exception, keep the port if it's a bond fake
719 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
720 && !strcmp(ofproto_port.type, "internal")) {
726 if (strcmp(ofproto_port.type, iface->type)
727 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
728 /* The interface is the wrong type or can't be configured.
733 /* If the requested OpenFlow port for 'iface' changed, and it's not
734 * already the correct port, then we might want to temporarily delete
735 * this interface, so we can add it back again with the new OpenFlow
737 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
738 if (iface->ofp_port != OFPP_LOCAL &&
739 requested_ofp_port != OFPP_NONE &&
740 requested_ofp_port != iface->ofp_port) {
741 ofp_port_t victim_request;
742 struct iface *victim;
744 /* Check for an existing OpenFlow port currently occupying
745 * 'iface''s requested port number. If there isn't one, then
746 * delete this port. Otherwise we need to consider further. */
747 victim = iface_from_ofp_port(br, requested_ofp_port);
752 /* 'victim' is a port currently using 'iface''s requested port
753 * number. Unless 'victim' specifically requested that port
754 * number, too, then we can delete both 'iface' and 'victim'
755 * temporarily. (We'll add both of them back again later with new
756 * OpenFlow port numbers.)
758 * If 'victim' did request port number 'requested_ofp_port', just
759 * like 'iface', then that's a configuration inconsistency that we
760 * can't resolve. We might as well let it keep its current port
762 victim_request = iface_get_requested_ofp_port(victim->cfg);
763 if (victim_request != requested_ofp_port) {
764 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
765 iface_destroy(victim);
774 iface_destroy(iface);
775 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
777 for (i = 0; i < n; i++) {
778 ofproto_port_del(br->ofproto, del[i]);
782 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
783 * that we didn't see when we iterated through the datapath, i.e. ports
784 * that disappeared underneath use. This is an unusual situation, but it
785 * can happen in some cases:
787 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
788 * idea but could happen).
790 * - The port represented a device that disappeared, e.g. a tuntap
791 * device destroyed via "tunctl -d", a physical Ethernet device
792 * whose module was just unloaded via "rmmod", or a virtual NIC for a
793 * VM whose VM was just terminated. */
794 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
795 struct iface *iface, *iface_next;
797 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
798 if (!sset_contains(&ofproto_ports, iface->name)) {
799 iface_destroy__(iface);
803 if (list_is_empty(&port->ifaces)) {
807 sset_destroy(&ofproto_ports);
811 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
812 bool with_requested_port)
814 struct shash_node *port_node;
816 SHASH_FOR_EACH (port_node, wanted_ports) {
817 const struct ovsrec_port *port_cfg = port_node->data;
820 for (i = 0; i < port_cfg->n_interfaces; i++) {
821 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
822 ofp_port_t requested_ofp_port;
824 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
825 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
826 struct iface *iface = iface_lookup(br, iface_cfg->name);
829 iface_create(br, iface_cfg, port_cfg);
837 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
839 /* First add interfaces that request a particular port number. */
840 bridge_add_ports__(br, wanted_ports, true);
842 /* Then add interfaces that want automatic port number assignment.
843 * We add these afterward to avoid accidentally taking a specifically
844 * requested port number. */
845 bridge_add_ports__(br, wanted_ports, false);
849 port_configure(struct port *port)
851 const struct ovsrec_port *cfg = port->cfg;
852 struct bond_settings bond_settings;
853 struct lacp_settings lacp_settings;
854 struct ofproto_bundle_settings s;
857 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
858 configure_splinter_port(port);
867 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
868 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
869 s.slaves[s.n_slaves++] = iface->ofp_port;
874 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
878 /* Get VLAN trunks. */
881 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
885 if (cfg->vlan_mode) {
886 if (!strcmp(cfg->vlan_mode, "access")) {
887 s.vlan_mode = PORT_VLAN_ACCESS;
888 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
889 s.vlan_mode = PORT_VLAN_TRUNK;
890 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
891 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
892 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
893 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
895 /* This "can't happen" because ovsdb-server should prevent it. */
896 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
897 "back to trunk mode", port->name, cfg->vlan_mode);
898 s.vlan_mode = PORT_VLAN_TRUNK;
902 s.vlan_mode = PORT_VLAN_ACCESS;
904 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
908 s.vlan_mode = PORT_VLAN_TRUNK;
911 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
914 /* Get LACP settings. */
915 s.lacp = port_configure_lacp(port, &lacp_settings);
919 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
920 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
921 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
924 s.lacp_slaves = NULL;
927 /* Get bond settings. */
928 if (s.n_slaves > 1) {
929 s.bond = &bond_settings;
930 port_configure_bond(port, &bond_settings);
933 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
934 netdev_set_miimon_interval(iface->netdev, 0);
939 ofproto_bundle_register(port->bridge->ofproto, port, &s);
947 /* Pick local port hardware address and datapath ID for 'br'. */
949 bridge_configure_datapath_id(struct bridge *br)
951 uint8_t ea[ETH_ADDR_LEN];
953 struct iface *local_iface;
954 struct iface *hw_addr_iface;
957 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
958 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
960 int error = netdev_set_etheraddr(local_iface->netdev, ea);
962 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
963 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
964 "Ethernet address: %s",
965 br->name, ovs_strerror(error));
968 memcpy(br->ea, ea, ETH_ADDR_LEN);
970 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
971 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
972 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
973 ofproto_set_datapath_id(br->ofproto, dpid);
976 dpid_string = xasprintf("%016"PRIx64, dpid);
977 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
981 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
984 bridge_get_allowed_versions(struct bridge *br)
986 uint32_t allowed_versions;
988 if (!br->cfg->n_protocols)
991 allowed_versions = ofputil_versions_from_strings(br->cfg->protocols,
992 br->cfg->n_protocols);
994 allowed_versions &= ~(1u << OFP14_VERSION);
996 return allowed_versions;
999 /* Set NetFlow configuration on 'br'. */
1001 bridge_configure_netflow(struct bridge *br)
1003 struct ovsrec_netflow *cfg = br->cfg->netflow;
1004 struct netflow_options opts;
1007 ofproto_set_netflow(br->ofproto, NULL);
1011 memset(&opts, 0, sizeof opts);
1013 /* Get default NetFlow configuration from datapath.
1014 * Apply overrides from 'cfg'. */
1015 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1016 if (cfg->engine_type) {
1017 opts.engine_type = *cfg->engine_type;
1019 if (cfg->engine_id) {
1020 opts.engine_id = *cfg->engine_id;
1023 /* Configure active timeout interval. */
1024 opts.active_timeout = cfg->active_timeout;
1025 if (!opts.active_timeout) {
1026 opts.active_timeout = -1;
1027 } else if (opts.active_timeout < 0) {
1028 VLOG_WARN("bridge %s: active timeout interval set to negative "
1029 "value, using default instead (%d seconds)", br->name,
1030 NF_ACTIVE_TIMEOUT_DEFAULT);
1031 opts.active_timeout = -1;
1034 /* Add engine ID to interface number to disambiguate bridgs? */
1035 opts.add_id_to_iface = cfg->add_id_to_interface;
1036 if (opts.add_id_to_iface) {
1037 if (opts.engine_id > 0x7f) {
1038 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1039 "another vswitch, choose an engine id less than 128",
1042 if (hmap_count(&br->ports) > 508) {
1043 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1044 "another port when more than 508 ports are used",
1050 sset_init(&opts.collectors);
1051 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1054 if (ofproto_set_netflow(br->ofproto, &opts)) {
1055 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1057 sset_destroy(&opts.collectors);
1060 /* Set sFlow configuration on 'br'. */
1062 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1064 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1065 struct ovsrec_controller **controllers;
1066 struct ofproto_sflow_options oso;
1067 size_t n_controllers;
1071 ofproto_set_sflow(br->ofproto, NULL);
1075 memset(&oso, 0, sizeof oso);
1077 sset_init(&oso.targets);
1078 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1080 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1081 if (cfg->sampling) {
1082 oso.sampling_rate = *cfg->sampling;
1085 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1087 oso.polling_interval = *cfg->polling;
1090 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1092 oso.header_len = *cfg->header;
1095 oso.sub_id = (*sflow_bridge_number)++;
1096 oso.agent_device = cfg->agent;
1098 oso.control_ip = NULL;
1099 n_controllers = bridge_get_controllers(br, &controllers);
1100 for (i = 0; i < n_controllers; i++) {
1101 if (controllers[i]->local_ip) {
1102 oso.control_ip = controllers[i]->local_ip;
1106 ofproto_set_sflow(br->ofproto, &oso);
1108 sset_destroy(&oso.targets);
1111 /* Returns whether a IPFIX row is valid. */
1113 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1115 return ipfix && ipfix->n_targets > 0;
1118 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1120 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1121 const struct bridge *br)
1123 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1126 /* Set IPFIX configuration on 'br'. */
1128 bridge_configure_ipfix(struct bridge *br)
1130 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1131 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1132 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1133 struct ofproto_ipfix_bridge_exporter_options be_opts;
1134 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1135 size_t n_fe_opts = 0;
1137 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1138 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1143 if (!valid_be_cfg && n_fe_opts == 0) {
1144 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1149 memset(&be_opts, 0, sizeof be_opts);
1151 sset_init(&be_opts.targets);
1152 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1154 if (be_cfg->sampling) {
1155 be_opts.sampling_rate = *be_cfg->sampling;
1157 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1159 if (be_cfg->obs_domain_id) {
1160 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1162 if (be_cfg->obs_point_id) {
1163 be_opts.obs_point_id = *be_cfg->obs_point_id;
1165 if (be_cfg->cache_active_timeout) {
1166 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1168 if (be_cfg->cache_max_flows) {
1169 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1173 if (n_fe_opts > 0) {
1174 struct ofproto_ipfix_flow_exporter_options *opts;
1175 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1177 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1178 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1179 opts->collector_set_id = fe_cfg->id;
1180 sset_init(&opts->targets);
1181 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1182 fe_cfg->ipfix->n_targets);
1183 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1184 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1185 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1186 ? *fe_cfg->ipfix->cache_max_flows : 0;
1192 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1196 sset_destroy(&be_opts.targets);
1199 if (n_fe_opts > 0) {
1200 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1202 for (i = 0; i < n_fe_opts; i++) {
1203 sset_destroy(&opts->targets);
1211 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1212 struct ofproto_port_stp_settings *port_s,
1213 int *port_num_counter, unsigned long *port_num_bitmap)
1215 const char *config_str;
1216 struct iface *iface;
1218 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1219 port_s->enable = false;
1222 port_s->enable = true;
1225 /* STP over bonds is not supported. */
1226 if (!list_is_singleton(&port->ifaces)) {
1227 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1229 port_s->enable = false;
1233 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1235 /* Internal ports shouldn't participate in spanning tree, so
1237 if (!strcmp(iface->type, "internal")) {
1238 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1239 port_s->enable = false;
1243 /* STP on mirror output ports is not supported. */
1244 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1245 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1246 port_s->enable = false;
1250 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1252 unsigned long int port_num = strtoul(config_str, NULL, 0);
1253 int port_idx = port_num - 1;
1255 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1256 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1257 port_s->enable = false;
1261 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1262 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1263 port->name, port_num);
1264 port_s->enable = false;
1267 bitmap_set1(port_num_bitmap, port_idx);
1268 port_s->port_num = port_idx;
1270 if (*port_num_counter >= STP_MAX_PORTS) {
1271 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1272 port_s->enable = false;
1276 port_s->port_num = (*port_num_counter)++;
1279 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1281 port_s->path_cost = strtoul(config_str, NULL, 10);
1283 enum netdev_features current;
1286 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1287 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1288 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1291 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1293 port_s->priority = strtoul(config_str, NULL, 0);
1295 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1299 /* Set spanning tree configuration on 'br'. */
1301 bridge_configure_stp(struct bridge *br)
1303 if (!br->cfg->stp_enable) {
1304 ofproto_set_stp(br->ofproto, NULL);
1306 struct ofproto_stp_settings br_s;
1307 const char *config_str;
1309 int port_num_counter;
1310 unsigned long *port_num_bitmap;
1312 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1314 uint8_t ea[ETH_ADDR_LEN];
1316 if (eth_addr_from_string(config_str, ea)) {
1317 br_s.system_id = eth_addr_to_uint64(ea);
1319 br_s.system_id = eth_addr_to_uint64(br->ea);
1320 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1321 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1324 br_s.system_id = eth_addr_to_uint64(br->ea);
1327 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1329 br_s.priority = strtoul(config_str, NULL, 0);
1331 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1334 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1336 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1338 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1341 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1343 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1345 br_s.max_age = STP_DEFAULT_MAX_AGE;
1348 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1350 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1352 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1355 /* Configure STP on the bridge. */
1356 if (ofproto_set_stp(br->ofproto, &br_s)) {
1357 VLOG_ERR("bridge %s: could not enable STP", br->name);
1361 /* Users must either set the port number with the "stp-port-num"
1362 * configuration on all ports or none. If manual configuration
1363 * is not done, then we allocate them sequentially. */
1364 port_num_counter = 0;
1365 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1366 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1367 struct ofproto_port_stp_settings port_s;
1368 struct iface *iface;
1370 port_configure_stp(br->ofproto, port, &port_s,
1371 &port_num_counter, port_num_bitmap);
1373 /* As bonds are not supported, just apply configuration to
1374 * all interfaces. */
1375 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1376 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1378 VLOG_ERR("port %s: could not enable STP", port->name);
1384 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1385 && port_num_counter) {
1386 VLOG_ERR("bridge %s: must manually configure all STP port "
1387 "IDs or none, disabling", br->name);
1388 ofproto_set_stp(br->ofproto, NULL);
1390 bitmap_free(port_num_bitmap);
1395 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1397 const struct port *port = port_lookup(br, name);
1398 return port && port_is_bond_fake_iface(port);
1402 port_is_bond_fake_iface(const struct port *port)
1404 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1408 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1410 struct bridge *br, *next;
1411 struct shash new_br;
1414 /* Collect new bridges' names and types. */
1415 shash_init(&new_br);
1416 for (i = 0; i < cfg->n_bridges; i++) {
1417 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1418 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1420 if (strchr(br_cfg->name, '/')) {
1421 /* Prevent remote ovsdb-server users from accessing arbitrary
1422 * directories, e.g. consider a bridge named "../../../etc/". */
1423 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1425 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1426 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1430 /* Get rid of deleted bridges or those whose types have changed.
1431 * Update 'cfg' of bridges that still exist. */
1432 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1433 br->cfg = shash_find_data(&new_br, br->name);
1434 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1435 br->cfg->datapath_type))) {
1440 /* Add new bridges. */
1441 for (i = 0; i < cfg->n_bridges; i++) {
1442 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1443 struct bridge *br = bridge_lookup(br_cfg->name);
1445 bridge_create(br_cfg);
1449 shash_destroy(&new_br);
1452 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1453 * Returns 0 if successful, otherwise a positive errno value. */
1455 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1456 struct netdev *netdev)
1458 return netdev_set_config(netdev, &iface_cfg->options);
1461 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1462 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1464 * If successful, returns 0 and stores the network device in '*netdevp'. On
1465 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1467 iface_do_create(const struct bridge *br,
1468 const struct ovsrec_interface *iface_cfg,
1469 const struct ovsrec_port *port_cfg,
1470 ofp_port_t *ofp_portp, struct netdev **netdevp)
1472 struct netdev *netdev = NULL;
1475 if (netdev_is_reserved_name(iface_cfg->name)) {
1476 VLOG_WARN("could not create interface %s, name is reserved",
1482 error = netdev_open(iface_cfg->name,
1483 iface_get_type(iface_cfg, br->cfg), &netdev);
1485 VLOG_WARN("could not open network device %s (%s)",
1486 iface_cfg->name, ovs_strerror(error));
1490 error = iface_set_netdev_config(iface_cfg, netdev);
1495 *ofp_portp = iface_pick_ofport(iface_cfg);
1496 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1501 VLOG_INFO("bridge %s: added interface %s on port %d",
1502 br->name, iface_cfg->name, *ofp_portp);
1504 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1505 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1513 netdev_close(netdev);
1517 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1518 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1519 * automatically allocated for the iface. Takes ownership of and
1520 * deallocates 'if_cfg'.
1522 * Return true if an iface is successfully created, false otherwise. */
1524 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1525 const struct ovsrec_port *port_cfg)
1527 struct netdev *netdev;
1528 struct iface *iface;
1529 ofp_port_t ofp_port;
1533 /* Do the bits that can fail up front. */
1534 ovs_assert(!iface_lookup(br, iface_cfg->name));
1535 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1537 iface_clear_db_record(iface_cfg);
1541 /* Get or create the port structure. */
1542 port = port_lookup(br, port_cfg->name);
1544 port = port_create(br, port_cfg);
1547 /* Create the iface structure. */
1548 iface = xzalloc(sizeof *iface);
1549 list_push_back(&port->ifaces, &iface->port_elem);
1550 hmap_insert(&br->iface_by_name, &iface->name_node,
1551 hash_string(iface_cfg->name, 0));
1553 iface->name = xstrdup(iface_cfg->name);
1554 iface->ofp_port = ofp_port;
1555 iface->netdev = netdev;
1556 iface->type = iface_get_type(iface_cfg, br->cfg);
1557 iface->cfg = iface_cfg;
1558 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1559 hash_ofp_port(ofp_port));
1561 /* Populate initial status in database. */
1562 iface_refresh_stats(iface);
1563 iface_refresh_netdev_status(iface);
1565 /* Add bond fake iface if necessary. */
1566 if (port_is_bond_fake_iface(port)) {
1567 struct ofproto_port ofproto_port;
1569 if (ofproto_port_query_by_name(br->ofproto, port->name,
1571 struct netdev *netdev;
1574 error = netdev_open(port->name, "internal", &netdev);
1576 ofp_port_t fake_ofp_port = OFPP_NONE;
1577 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1578 netdev_close(netdev);
1580 VLOG_WARN("could not open network device %s (%s)",
1581 port->name, ovs_strerror(error));
1584 /* Already exists, nothing to do. */
1585 ofproto_port_destroy(&ofproto_port);
1592 /* Set forward BPDU option. */
1594 bridge_configure_forward_bpdu(struct bridge *br)
1596 ofproto_set_forward_bpdu(br->ofproto,
1597 smap_get_bool(&br->cfg->other_config,
1602 /* Set MAC learning table configuration for 'br'. */
1604 bridge_configure_mac_table(struct bridge *br)
1606 const char *idle_time_str;
1609 const char *mac_table_size_str;
1612 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1613 idle_time = (idle_time_str && atoi(idle_time_str)
1614 ? atoi(idle_time_str)
1615 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1617 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1618 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1619 ? atoi(mac_table_size_str)
1622 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1626 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1627 const struct port *fake_br, struct iface **hw_addr_iface)
1629 struct hmapx mirror_output_ports;
1631 bool found_addr = false;
1635 /* Mirror output ports don't participate in picking the local hardware
1636 * address. ofproto can't help us find out whether a given port is a
1637 * mirror output because we haven't configured mirrors yet, so we need to
1638 * accumulate them ourselves. */
1639 hmapx_init(&mirror_output_ports);
1640 for (i = 0; i < br->cfg->n_mirrors; i++) {
1641 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1642 if (m->output_port) {
1643 hmapx_add(&mirror_output_ports, m->output_port);
1647 /* Otherwise choose the minimum non-local MAC address among all of the
1649 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1650 uint8_t iface_ea[ETH_ADDR_LEN];
1651 struct iface *candidate;
1652 struct iface *iface;
1654 /* Mirror output ports don't participate. */
1655 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1659 /* Choose the MAC address to represent the port. */
1661 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1662 /* Find the interface with this Ethernet address (if any) so that
1663 * we can provide the correct devname to the caller. */
1664 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1665 uint8_t candidate_ea[ETH_ADDR_LEN];
1666 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1667 && eth_addr_equals(iface_ea, candidate_ea)) {
1672 /* Choose the interface whose MAC address will represent the port.
1673 * The Linux kernel bonding code always chooses the MAC address of
1674 * the first slave added to a bond, and the Fedora networking
1675 * scripts always add slaves to a bond in alphabetical order, so
1676 * for compatibility we choose the interface with the name that is
1677 * first in alphabetical order. */
1678 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1679 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1684 /* The local port doesn't count (since we're trying to choose its
1685 * MAC address anyway). */
1686 if (iface->ofp_port == OFPP_LOCAL) {
1690 /* For fake bridges we only choose from ports with the same tag */
1691 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1692 if (!port->cfg->tag) {
1695 if (*port->cfg->tag != *fake_br->cfg->tag) {
1701 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1707 /* Compare against our current choice. */
1708 if (!eth_addr_is_multicast(iface_ea) &&
1709 !eth_addr_is_local(iface_ea) &&
1710 !eth_addr_is_reserved(iface_ea) &&
1711 !eth_addr_is_zero(iface_ea) &&
1712 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1714 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1715 *hw_addr_iface = iface;
1721 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1722 *hw_addr_iface = NULL;
1725 hmapx_destroy(&mirror_output_ports);
1729 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1730 struct iface **hw_addr_iface)
1733 *hw_addr_iface = NULL;
1735 /* Did the user request a particular MAC? */
1736 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1737 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1738 if (eth_addr_is_multicast(ea)) {
1739 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1740 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1741 } else if (eth_addr_is_zero(ea)) {
1742 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1748 /* Find a local hw address */
1749 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1752 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1753 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1754 * an interface on 'br', then that interface must be passed in as
1755 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1756 * 'hw_addr_iface' must be passed in as a null pointer. */
1758 bridge_pick_datapath_id(struct bridge *br,
1759 const uint8_t bridge_ea[ETH_ADDR_LEN],
1760 struct iface *hw_addr_iface)
1763 * The procedure for choosing a bridge MAC address will, in the most
1764 * ordinary case, also choose a unique MAC that we can use as a datapath
1765 * ID. In some special cases, though, multiple bridges will end up with
1766 * the same MAC address. This is OK for the bridges, but it will confuse
1767 * the OpenFlow controller, because each datapath needs a unique datapath
1770 * Datapath IDs must be unique. It is also very desirable that they be
1771 * stable from one run to the next, so that policy set on a datapath
1774 const char *datapath_id;
1777 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1778 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1782 if (!hw_addr_iface) {
1784 * A purely internal bridge, that is, one that has no non-virtual
1785 * network devices on it at all, is difficult because it has no
1786 * natural unique identifier at all.
1788 * When the host is a XenServer, we handle this case by hashing the
1789 * host's UUID with the name of the bridge. Names of bridges are
1790 * persistent across XenServer reboots, although they can be reused if
1791 * an internal network is destroyed and then a new one is later
1792 * created, so this is fairly effective.
1794 * When the host is not a XenServer, we punt by using a random MAC
1795 * address on each run.
1797 const char *host_uuid = xenserver_get_host_uuid();
1799 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1800 dpid = dpid_from_hash(combined, strlen(combined));
1806 return eth_addr_to_uint64(bridge_ea);
1810 dpid_from_hash(const void *data, size_t n)
1812 uint8_t hash[SHA1_DIGEST_SIZE];
1814 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1815 sha1_bytes(data, n, hash);
1816 eth_addr_mark_random(hash);
1817 return eth_addr_to_uint64(hash);
1821 iface_refresh_netdev_status(struct iface *iface)
1825 enum netdev_features current;
1826 enum netdev_flags flags;
1827 const char *link_state;
1828 uint8_t mac[ETH_ADDR_LEN];
1829 int64_t bps, mtu_64, ifindex64, link_resets;
1832 if (iface_is_synthetic(iface)) {
1836 if (iface->change_seq == netdev_get_change_seq(iface->netdev)) {
1840 iface->change_seq = netdev_get_change_seq(iface->netdev);
1844 if (!netdev_get_status(iface->netdev, &smap)) {
1845 ovsrec_interface_set_status(iface->cfg, &smap);
1847 ovsrec_interface_set_status(iface->cfg, NULL);
1850 smap_destroy(&smap);
1852 error = netdev_get_flags(iface->netdev, &flags);
1854 const char *state = flags & NETDEV_UP ? "up" : "down";
1856 ovsrec_interface_set_admin_state(iface->cfg, state);
1858 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1861 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1862 ovsrec_interface_set_link_state(iface->cfg, link_state);
1864 link_resets = netdev_get_carrier_resets(iface->netdev);
1865 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1867 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1868 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1870 ovsrec_interface_set_duplex(iface->cfg,
1871 netdev_features_is_full_duplex(current)
1873 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1875 ovsrec_interface_set_duplex(iface->cfg, NULL);
1876 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1879 error = netdev_get_mtu(iface->netdev, &mtu);
1882 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1884 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1887 error = netdev_get_etheraddr(iface->netdev, mac);
1889 char mac_string[32];
1891 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1892 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1894 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1897 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1898 * if there is no valid ifindex number. */
1899 ifindex64 = netdev_get_ifindex(iface->netdev);
1900 if (ifindex64 < 0) {
1903 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1907 iface_refresh_ofproto_status(struct iface *iface)
1912 if (iface_is_synthetic(iface)) {
1916 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1920 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1922 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1925 iface_refresh_cfm_stats(iface);
1928 error = ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1929 iface->ofp_port, &smap);
1931 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1933 smap_destroy(&smap);
1936 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1939 iface_refresh_cfm_stats(struct iface *iface)
1941 const struct ovsrec_interface *cfg = iface->cfg;
1942 struct ofproto_cfm_status status;
1945 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1946 iface->ofp_port, &status);
1948 /* Do nothing if there is no status change since last update. */
1949 } else if (error > 0) {
1950 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1951 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1952 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1953 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1954 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1955 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1957 const char *reasons[CFM_FAULT_N_REASONS];
1958 int64_t cfm_health = status.health;
1959 int64_t cfm_flap_count = status.flap_count;
1960 bool faulted = status.faults != 0;
1963 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1966 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1967 int reason = 1 << i;
1968 if (status.faults & reason) {
1969 reasons[j++] = cfm_fault_reason_to_str(reason);
1972 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1974 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1976 if (status.remote_opstate >= 0) {
1977 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1978 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1980 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1983 ovsrec_interface_set_cfm_remote_mpids(cfg,
1984 (const int64_t *)status.rmps,
1986 if (cfm_health >= 0) {
1987 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1989 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1997 iface_refresh_stats(struct iface *iface)
1999 #define IFACE_STATS \
2000 IFACE_STAT(rx_packets, "rx_packets") \
2001 IFACE_STAT(tx_packets, "tx_packets") \
2002 IFACE_STAT(rx_bytes, "rx_bytes") \
2003 IFACE_STAT(tx_bytes, "tx_bytes") \
2004 IFACE_STAT(rx_dropped, "rx_dropped") \
2005 IFACE_STAT(tx_dropped, "tx_dropped") \
2006 IFACE_STAT(rx_errors, "rx_errors") \
2007 IFACE_STAT(tx_errors, "tx_errors") \
2008 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2009 IFACE_STAT(rx_over_errors, "rx_over_err") \
2010 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2011 IFACE_STAT(collisions, "collisions")
2013 #define IFACE_STAT(MEMBER, NAME) + 1
2014 enum { N_IFACE_STATS = IFACE_STATS };
2016 int64_t values[N_IFACE_STATS];
2017 char *keys[N_IFACE_STATS];
2020 struct netdev_stats stats;
2022 if (iface_is_synthetic(iface)) {
2026 /* Intentionally ignore return value, since errors will set 'stats' to
2027 * all-1s, and we will deal with that correctly below. */
2028 netdev_get_stats(iface->netdev, &stats);
2030 /* Copy statistics into keys[] and values[]. */
2032 #define IFACE_STAT(MEMBER, NAME) \
2033 if (stats.MEMBER != UINT64_MAX) { \
2035 values[n] = stats.MEMBER; \
2040 ovs_assert(n <= N_IFACE_STATS);
2042 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2047 br_refresh_stp_status(struct bridge *br)
2049 struct smap smap = SMAP_INITIALIZER(&smap);
2050 struct ofproto *ofproto = br->ofproto;
2051 struct ofproto_stp_status status;
2053 if (ofproto_get_stp_status(ofproto, &status)) {
2057 if (!status.enabled) {
2058 ovsrec_bridge_set_status(br->cfg, NULL);
2062 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2063 STP_ID_ARGS(status.bridge_id));
2064 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2065 STP_ID_ARGS(status.designated_root));
2066 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2068 ovsrec_bridge_set_status(br->cfg, &smap);
2069 smap_destroy(&smap);
2073 port_refresh_stp_status(struct port *port)
2075 struct ofproto *ofproto = port->bridge->ofproto;
2076 struct iface *iface;
2077 struct ofproto_port_stp_status status;
2080 if (port_is_synthetic(port)) {
2084 /* STP doesn't currently support bonds. */
2085 if (!list_is_singleton(&port->ifaces)) {
2086 ovsrec_port_set_status(port->cfg, NULL);
2090 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2091 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2095 if (!status.enabled) {
2096 ovsrec_port_set_status(port->cfg, NULL);
2100 /* Set Status column. */
2102 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2103 smap_add(&smap, "stp_state", stp_state_name(status.state));
2104 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2105 smap_add(&smap, "stp_role", stp_role_name(status.role));
2106 ovsrec_port_set_status(port->cfg, &smap);
2107 smap_destroy(&smap);
2111 port_refresh_stp_stats(struct port *port)
2113 struct ofproto *ofproto = port->bridge->ofproto;
2114 struct iface *iface;
2115 struct ofproto_port_stp_stats stats;
2117 int64_t int_values[3];
2119 if (port_is_synthetic(port)) {
2123 /* STP doesn't currently support bonds. */
2124 if (!list_is_singleton(&port->ifaces)) {
2128 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2129 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2133 if (!stats.enabled) {
2134 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2138 /* Set Statistics column. */
2139 keys[0] = "stp_tx_count";
2140 int_values[0] = stats.tx_count;
2141 keys[1] = "stp_rx_count";
2142 int_values[1] = stats.rx_count;
2143 keys[2] = "stp_error_count";
2144 int_values[2] = stats.error_count;
2146 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2147 ARRAY_SIZE(int_values));
2151 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2153 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2157 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2159 bool enable = enable_system_stats(cfg);
2161 system_stats_enable(enable);
2163 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2168 run_system_stats(void)
2170 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2173 stats = system_stats_run();
2175 struct ovsdb_idl_txn *txn;
2176 struct ovsdb_datum datum;
2178 txn = ovsdb_idl_txn_create(idl);
2179 ovsdb_datum_from_smap(&datum, stats);
2180 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2182 ovsdb_idl_txn_commit(txn);
2183 ovsdb_idl_txn_destroy(txn);
2190 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2193 case OFPCR12_ROLE_EQUAL:
2195 case OFPCR12_ROLE_MASTER:
2197 case OFPCR12_ROLE_SLAVE:
2199 case OFPCR12_ROLE_NOCHANGE:
2201 return "*** INVALID ROLE ***";
2206 refresh_controller_status(void)
2210 const struct ovsrec_controller *cfg;
2214 /* Accumulate status for controllers on all bridges. */
2215 HMAP_FOR_EACH (br, node, &all_bridges) {
2216 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2219 /* Update each controller in the database with current status. */
2220 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2221 struct ofproto_controller_info *cinfo =
2222 shash_find_data(&info, cfg->target);
2225 struct smap smap = SMAP_INITIALIZER(&smap);
2226 const char **values = cinfo->pairs.values;
2227 const char **keys = cinfo->pairs.keys;
2230 for (i = 0; i < cinfo->pairs.n; i++) {
2231 smap_add(&smap, keys[i], values[i]);
2234 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2235 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2237 ovsrec_controller_set_status(cfg, &smap);
2238 smap_destroy(&smap);
2240 ovsrec_controller_set_is_connected(cfg, false);
2241 ovsrec_controller_set_role(cfg, NULL);
2242 ovsrec_controller_set_status(cfg, NULL);
2246 ofproto_free_ofproto_controller_info(&info);
2256 /* Let each datapath type do the work that it needs to do. */
2258 ofproto_enumerate_types(&types);
2259 SSET_FOR_EACH (type, &types) {
2260 ofproto_type_run(type);
2262 sset_destroy(&types);
2264 /* Let each bridge do the work that it needs to do. */
2265 HMAP_FOR_EACH (br, node, &all_bridges) {
2266 ofproto_run(br->ofproto);
2273 static struct ovsrec_open_vswitch null_cfg;
2274 const struct ovsrec_open_vswitch *cfg;
2276 bool vlan_splinters_changed;
2279 ovsrec_open_vswitch_init(&null_cfg);
2283 if (ovsdb_idl_is_lock_contended(idl)) {
2284 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2285 struct bridge *br, *next_br;
2287 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2288 "disabling this process (pid %ld) until it goes away",
2289 (long int) getpid());
2291 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2294 /* Since we will not be running system_stats_run() in this process
2295 * with the current situation of multiple ovs-vswitchd daemons,
2296 * disable system stats collection. */
2297 system_stats_enable(false);
2299 } else if (!ovsdb_idl_has_lock(idl)) {
2302 cfg = ovsrec_open_vswitch_first(idl);
2304 /* Initialize the ofproto library. This only needs to run once, but
2305 * it must be done after the configuration is set. If the
2306 * initialization has already occurred, bridge_init_ofproto()
2307 * returns immediately. */
2308 bridge_init_ofproto(cfg);
2310 /* Once the value of flow-restore-wait is false, we no longer should
2311 * check its value from the database. */
2312 if (cfg && ofproto_get_flow_restore_wait()) {
2313 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2314 "flow-restore-wait", false));
2319 /* Re-configure SSL. We do this on every trip through the main loop,
2320 * instead of just when the database changes, because the contents of the
2321 * key and certificate files can change without the database changing.
2323 * We do this before bridge_reconfigure() because that function might
2324 * initiate SSL connections and thus requires SSL to be configured. */
2325 if (cfg && cfg->ssl) {
2326 const struct ovsrec_ssl *ssl = cfg->ssl;
2328 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2329 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2332 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2333 * usage has changed. */
2334 vlan_splinters_changed = false;
2335 if (vlan_splinters_enabled_anywhere) {
2336 HMAP_FOR_EACH (br, node, &all_bridges) {
2337 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2338 vlan_splinters_changed = true;
2344 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2345 struct ovsdb_idl_txn *txn;
2347 idl_seqno = ovsdb_idl_get_seqno(idl);
2348 txn = ovsdb_idl_txn_create(idl);
2349 bridge_reconfigure(cfg ? cfg : &null_cfg);
2352 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2355 /* If we are completing our initial configuration for this run
2356 * of ovs-vswitchd, then keep the transaction around to monitor
2357 * it for completion. */
2358 if (initial_config_done) {
2359 ovsdb_idl_txn_commit(txn);
2360 ovsdb_idl_txn_destroy(txn);
2362 initial_config_done = true;
2363 daemonize_txn = txn;
2367 if (daemonize_txn) {
2368 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2369 if (status != TXN_INCOMPLETE) {
2370 ovsdb_idl_txn_destroy(daemonize_txn);
2371 daemonize_txn = NULL;
2373 /* ovs-vswitchd has completed initialization, so allow the
2374 * process that forked us to exit successfully. */
2375 daemonize_complete();
2377 vlog_enable_async();
2379 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2383 /* Refresh interface and mirror stats if necessary. */
2384 if (time_msec() >= iface_stats_timer) {
2386 struct ovsdb_idl_txn *txn;
2388 txn = ovsdb_idl_txn_create(idl);
2389 HMAP_FOR_EACH (br, node, &all_bridges) {
2393 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2394 struct iface *iface;
2396 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2397 iface_refresh_stats(iface);
2400 port_refresh_stp_stats(port);
2403 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2404 mirror_refresh_stats(m);
2408 refresh_controller_status();
2409 ovsdb_idl_txn_commit(txn);
2410 ovsdb_idl_txn_destroy(txn); /* XXX */
2413 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2419 /* Check the need to update status. */
2420 seq = seq_read(connectivity_seq_get());
2421 if (seq != connectivity_seqno) {
2422 connectivity_seqno = seq;
2423 status_txn = ovsdb_idl_txn_create(idl);
2424 HMAP_FOR_EACH (br, node, &all_bridges) {
2427 br_refresh_stp_status(br);
2428 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2429 struct iface *iface;
2431 port_refresh_stp_status(port);
2432 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2433 iface_refresh_netdev_status(iface);
2434 iface_refresh_ofproto_status(iface);
2442 enum ovsdb_idl_txn_status status;
2444 status = ovsdb_idl_txn_commit(status_txn);
2445 /* Do not destroy "status_txn" if the transaction is
2446 * "TXN_INCOMPLETE". */
2447 if (status != TXN_INCOMPLETE) {
2448 ovsdb_idl_txn_destroy(status_txn);
2462 ovsdb_idl_wait(idl);
2463 if (daemonize_txn) {
2464 ovsdb_idl_txn_wait(daemonize_txn);
2468 ofproto_enumerate_types(&types);
2469 SSET_FOR_EACH (type, &types) {
2470 ofproto_type_wait(type);
2472 sset_destroy(&types);
2474 if (!hmap_is_empty(&all_bridges)) {
2477 HMAP_FOR_EACH (br, node, &all_bridges) {
2478 ofproto_wait(br->ofproto);
2480 poll_timer_wait_until(iface_stats_timer);
2483 /* If the status database transaction is 'TXN_INCOMPLETE' in this run,
2484 * register a timeout in 'STATUS_CHECK_AGAIN_MSEC'. Else, wait on the
2485 * global connectivity sequence number. Note, this also helps batch
2486 * multiple status changes into one transaction. */
2488 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2490 seq_wait(connectivity_seq_get(), connectivity_seqno);
2493 system_stats_wait();
2496 /* Adds some memory usage statistics for bridges into 'usage', for use with
2497 * memory_report(). */
2499 bridge_get_memory_usage(struct simap *usage)
2506 ofproto_enumerate_types(&types);
2507 SSET_FOR_EACH (type, &types) {
2508 ofproto_type_get_memory_usage(type, usage);
2510 sset_destroy(&types);
2512 HMAP_FOR_EACH (br, node, &all_bridges) {
2513 ofproto_get_memory_usage(br->ofproto, usage);
2517 /* QoS unixctl user interface functions. */
2519 struct qos_unixctl_show_cbdata {
2521 struct iface *iface;
2525 qos_unixctl_show_queue(unsigned int queue_id,
2526 const struct smap *details,
2527 struct iface *iface,
2530 struct netdev_queue_stats stats;
2531 struct smap_node *node;
2534 ds_put_cstr(ds, "\n");
2536 ds_put_format(ds, "Queue %u:\n", queue_id);
2538 ds_put_cstr(ds, "Default:\n");
2541 SMAP_FOR_EACH (node, details) {
2542 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2545 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2547 if (stats.tx_packets != UINT64_MAX) {
2548 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2551 if (stats.tx_bytes != UINT64_MAX) {
2552 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2555 if (stats.tx_errors != UINT64_MAX) {
2556 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2559 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2560 queue_id, ovs_strerror(error));
2565 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2566 const char *argv[], void *aux OVS_UNUSED)
2568 struct ds ds = DS_EMPTY_INITIALIZER;
2569 struct smap smap = SMAP_INITIALIZER(&smap);
2570 struct iface *iface;
2572 struct smap_node *node;
2574 iface = iface_find(argv[1]);
2576 unixctl_command_reply_error(conn, "no such interface");
2580 netdev_get_qos(iface->netdev, &type, &smap);
2582 if (*type != '\0') {
2583 struct netdev_queue_dump dump;
2584 struct smap details;
2585 unsigned int queue_id;
2587 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2589 SMAP_FOR_EACH (node, &smap) {
2590 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2593 smap_init(&details);
2594 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2595 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2597 smap_destroy(&details);
2599 unixctl_command_reply(conn, ds_cstr(&ds));
2601 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2602 unixctl_command_reply_error(conn, ds_cstr(&ds));
2605 smap_destroy(&smap);
2609 /* Bridge reconfiguration functions. */
2611 bridge_create(const struct ovsrec_bridge *br_cfg)
2615 ovs_assert(!bridge_lookup(br_cfg->name));
2616 br = xzalloc(sizeof *br);
2618 br->name = xstrdup(br_cfg->name);
2619 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2622 /* Derive the default Ethernet address from the bridge's UUID. This should
2623 * be unique and it will be stable between ovs-vswitchd runs. */
2624 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2625 eth_addr_mark_random(br->default_ea);
2627 hmap_init(&br->ports);
2628 hmap_init(&br->ifaces);
2629 hmap_init(&br->iface_by_name);
2630 hmap_init(&br->mirrors);
2632 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2636 bridge_destroy(struct bridge *br)
2639 struct mirror *mirror, *next_mirror;
2640 struct port *port, *next_port;
2642 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2645 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2646 mirror_destroy(mirror);
2649 hmap_remove(&all_bridges, &br->node);
2650 ofproto_destroy(br->ofproto);
2651 hmap_destroy(&br->ifaces);
2652 hmap_destroy(&br->ports);
2653 hmap_destroy(&br->iface_by_name);
2654 hmap_destroy(&br->mirrors);
2661 static struct bridge *
2662 bridge_lookup(const char *name)
2666 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2667 if (!strcmp(br->name, name)) {
2674 /* Handle requests for a listing of all flows known by the OpenFlow
2675 * stack, including those normally hidden. */
2677 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2678 const char *argv[], void *aux OVS_UNUSED)
2683 br = bridge_lookup(argv[1]);
2685 unixctl_command_reply_error(conn, "Unknown bridge");
2690 ofproto_get_all_flows(br->ofproto, &results);
2692 unixctl_command_reply(conn, ds_cstr(&results));
2693 ds_destroy(&results);
2696 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2697 * connections and reconnect. If BRIDGE is not specified, then all bridges
2698 * drop their controller connections and reconnect. */
2700 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2701 const char *argv[], void *aux OVS_UNUSED)
2705 br = bridge_lookup(argv[1]);
2707 unixctl_command_reply_error(conn, "Unknown bridge");
2710 ofproto_reconnect_controllers(br->ofproto);
2712 HMAP_FOR_EACH (br, node, &all_bridges) {
2713 ofproto_reconnect_controllers(br->ofproto);
2716 unixctl_command_reply(conn, NULL);
2720 bridge_get_controllers(const struct bridge *br,
2721 struct ovsrec_controller ***controllersp)
2723 struct ovsrec_controller **controllers;
2724 size_t n_controllers;
2726 controllers = br->cfg->controller;
2727 n_controllers = br->cfg->n_controller;
2729 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2735 *controllersp = controllers;
2737 return n_controllers;
2741 bridge_collect_wanted_ports(struct bridge *br,
2742 const unsigned long int *splinter_vlans,
2743 struct shash *wanted_ports)
2747 shash_init(wanted_ports);
2749 for (i = 0; i < br->cfg->n_ports; i++) {
2750 const char *name = br->cfg->ports[i]->name;
2751 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2752 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2756 if (bridge_get_controllers(br, NULL)
2757 && !shash_find(wanted_ports, br->name)) {
2758 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2759 br->name, br->name);
2761 ovsrec_interface_init(&br->synth_local_iface);
2762 ovsrec_port_init(&br->synth_local_port);
2764 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2765 br->synth_local_port.n_interfaces = 1;
2766 br->synth_local_port.name = br->name;
2768 br->synth_local_iface.name = br->name;
2769 br->synth_local_iface.type = "internal";
2771 br->synth_local_ifacep = &br->synth_local_iface;
2773 shash_add(wanted_ports, br->name, &br->synth_local_port);
2776 if (splinter_vlans) {
2777 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2781 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2782 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2783 * 'br' needs to complete its configuration. */
2785 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2787 struct shash_node *port_node;
2788 struct port *port, *next;
2790 /* Get rid of deleted ports.
2791 * Get rid of deleted interfaces on ports that still exist. */
2792 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2793 port->cfg = shash_find_data(wanted_ports, port->name);
2797 port_del_ifaces(port);
2801 /* Update iface->cfg and iface->type in interfaces that still exist. */
2802 SHASH_FOR_EACH (port_node, wanted_ports) {
2803 const struct ovsrec_port *port = port_node->data;
2806 for (i = 0; i < port->n_interfaces; i++) {
2807 const struct ovsrec_interface *cfg = port->interfaces[i];
2808 struct iface *iface = iface_lookup(br, cfg->name);
2809 const char *type = iface_get_type(cfg, br->cfg);
2814 } else if (!strcmp(type, "null")) {
2815 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2816 " may be removed in February 2013. Please email"
2817 " dev@openvswitch.org with concerns.",
2820 /* We will add new interfaces later. */
2826 /* Initializes 'oc' appropriately as a management service controller for
2829 * The caller must free oc->target when it is no longer needed. */
2831 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2832 struct ofproto_controller *oc)
2834 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2835 oc->max_backoff = 0;
2836 oc->probe_interval = 60;
2837 oc->band = OFPROTO_OUT_OF_BAND;
2839 oc->burst_limit = 0;
2840 oc->enable_async_msgs = true;
2843 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2845 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2846 struct ofproto_controller *oc)
2850 oc->target = c->target;
2851 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2852 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2853 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2854 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2855 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2856 oc->burst_limit = (c->controller_burst_limit
2857 ? *c->controller_burst_limit : 0);
2858 oc->enable_async_msgs = (!c->enable_async_messages
2859 || *c->enable_async_messages);
2860 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2861 if (dscp < 0 || dscp > 63) {
2862 dscp = DSCP_DEFAULT;
2867 /* Configures the IP stack for 'br''s local interface properly according to the
2868 * configuration in 'c'. */
2870 bridge_configure_local_iface_netdev(struct bridge *br,
2871 struct ovsrec_controller *c)
2873 struct netdev *netdev;
2874 struct in_addr mask, gateway;
2876 struct iface *local_iface;
2879 /* If there's no local interface or no IP address, give up. */
2880 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2881 if (!local_iface || !c->local_ip
2882 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2886 /* Bring up the local interface. */
2887 netdev = local_iface->netdev;
2888 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2890 /* Configure the IP address and netmask. */
2891 if (!c->local_netmask
2892 || !inet_pton(AF_INET, c->local_netmask, &mask)
2894 mask.s_addr = guess_netmask(ip.s_addr);
2896 if (!netdev_set_in4(netdev, ip, mask)) {
2897 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2898 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2901 /* Configure the default gateway. */
2902 if (c->local_gateway
2903 && inet_pton(AF_INET, c->local_gateway, &gateway)
2904 && gateway.s_addr) {
2905 if (!netdev_add_router(netdev, gateway)) {
2906 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2907 br->name, IP_ARGS(gateway.s_addr));
2912 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2913 * in either string are treated as equal to any number of slashes in the other,
2914 * e.g. "x///y" is equal to "x/y".
2916 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2917 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2918 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2919 * 'b' against a prefix of 'a'.
2922 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2924 const char *b_start = b;
2926 if (b - b_start >= b_stoplen) {
2928 } else if (*a != *b) {
2930 } else if (*a == '/') {
2931 a += strspn(a, "/");
2932 b += strspn(b, "/");
2933 } else if (*a == '\0') {
2943 bridge_configure_remotes(struct bridge *br,
2944 const struct sockaddr_in *managers, size_t n_managers)
2946 bool disable_in_band;
2948 struct ovsrec_controller **controllers;
2949 size_t n_controllers;
2951 enum ofproto_fail_mode fail_mode;
2953 struct ofproto_controller *ocs;
2957 /* Check if we should disable in-band control on this bridge. */
2958 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2961 /* Set OpenFlow queue ID for in-band control. */
2962 ofproto_set_in_band_queue(br->ofproto,
2963 smap_get_int(&br->cfg->other_config,
2964 "in-band-queue", -1));
2966 if (disable_in_band) {
2967 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2969 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2972 n_controllers = bridge_get_controllers(br, &controllers);
2974 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2977 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2978 for (i = 0; i < n_controllers; i++) {
2979 struct ovsrec_controller *c = controllers[i];
2981 if (!strncmp(c->target, "punix:", 6)
2982 || !strncmp(c->target, "unix:", 5)) {
2983 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2986 if (!strncmp(c->target, "unix:", 5)) {
2987 /* Connect to a listening socket */
2988 whitelist = xasprintf("unix:%s/", ovs_rundir());
2989 if (strchr(c->target, '/') &&
2990 !equal_pathnames(c->target, whitelist,
2991 strlen(whitelist))) {
2992 /* Absolute path specified, but not in ovs_rundir */
2993 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2994 "controller \"%s\" due to possibility for "
2995 "remote exploit. Instead, specify socket "
2996 "in whitelisted \"%s\" or connect to "
2997 "\"unix:%s/%s.mgmt\" (which is always "
2998 "available without special configuration).",
2999 br->name, c->target, whitelist,
3000 ovs_rundir(), br->name);
3005 whitelist = xasprintf("punix:%s/%s.controller",
3006 ovs_rundir(), br->name);
3007 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3008 /* Prevent remote ovsdb-server users from accessing
3009 * arbitrary Unix domain sockets and overwriting arbitrary
3011 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3012 "controller \"%s\" due to possibility of "
3013 "overwriting local files. Instead, specify "
3014 "whitelisted \"%s\" or connect to "
3015 "\"unix:%s/%s.mgmt\" (which is always "
3016 "available without special configuration).",
3017 br->name, c->target, whitelist,
3018 ovs_rundir(), br->name);
3027 bridge_configure_local_iface_netdev(br, c);
3028 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3029 if (disable_in_band) {
3030 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3035 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3036 bridge_get_allowed_versions(br));
3037 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3040 /* Set the fail-mode. */
3041 fail_mode = !br->cfg->fail_mode
3042 || !strcmp(br->cfg->fail_mode, "standalone")
3043 ? OFPROTO_FAIL_STANDALONE
3044 : OFPROTO_FAIL_SECURE;
3045 ofproto_set_fail_mode(br->ofproto, fail_mode);
3047 /* Configure OpenFlow controller connection snooping. */
3048 if (!ofproto_has_snoops(br->ofproto)) {
3052 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3053 ovs_rundir(), br->name));
3054 ofproto_set_snoops(br->ofproto, &snoops);
3055 sset_destroy(&snoops);
3060 bridge_configure_tables(struct bridge *br)
3062 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3066 n_tables = ofproto_get_n_tables(br->ofproto);
3068 for (i = 0; i < n_tables; i++) {
3069 struct ofproto_table_settings s;
3072 s.max_flows = UINT_MAX;
3075 s.n_prefix_fields = 0;
3076 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3078 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3079 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3082 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3083 s.max_flows = *cfg->flow_limit;
3085 if (cfg->overflow_policy
3086 && !strcmp(cfg->overflow_policy, "evict")) {
3088 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3089 for (k = 0; k < cfg->n_groups; k++) {
3090 const char *string = cfg->groups[k];
3093 msg = mf_parse_subfield__(&s.groups[k], &string);
3095 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3096 "'groups' (%s)", br->name, i, msg);
3098 } else if (*string) {
3099 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3100 "element '%s' contains trailing garbage",
3101 br->name, i, cfg->groups[k]);
3107 /* Prefix lookup fields. */
3108 s.n_prefix_fields = 0;
3109 for (k = 0; k < cfg->n_prefixes; k++) {
3110 const char *name = cfg->prefixes[k];
3111 const struct mf_field *mf = mf_from_name(name);
3113 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3117 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3118 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3119 "%s", br->name, name);
3122 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3123 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3124 "field not used: %s", br->name, name);
3127 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3129 if (s.n_prefix_fields > 0) {
3131 struct ds ds = DS_EMPTY_INITIALIZER;
3132 for (k = 0; k < s.n_prefix_fields; k++) {
3134 ds_put_char(&ds, ',');
3136 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3138 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3139 br->name, i, ds_cstr(&ds));
3144 ofproto_configure_table(br->ofproto, i, &s);
3148 for (; j < br->cfg->n_flow_tables; j++) {
3149 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3150 "%"PRId64" not supported by this datapath", br->name,
3151 br->cfg->key_flow_tables[j]);
3156 bridge_configure_dp_desc(struct bridge *br)
3158 ofproto_set_dp_desc(br->ofproto,
3159 smap_get(&br->cfg->other_config, "dp-desc"));
3162 /* Port functions. */
3164 static struct port *
3165 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3169 port = xzalloc(sizeof *port);
3171 port->name = xstrdup(cfg->name);
3173 list_init(&port->ifaces);
3175 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3179 /* Deletes interfaces from 'port' that are no longer configured for it. */
3181 port_del_ifaces(struct port *port)
3183 struct iface *iface, *next;
3184 struct sset new_ifaces;
3187 /* Collect list of new interfaces. */
3188 sset_init(&new_ifaces);
3189 for (i = 0; i < port->cfg->n_interfaces; i++) {
3190 const char *name = port->cfg->interfaces[i]->name;
3191 const char *type = port->cfg->interfaces[i]->type;
3192 if (strcmp(type, "null")) {
3193 sset_add(&new_ifaces, name);
3197 /* Get rid of deleted interfaces. */
3198 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3199 if (!sset_contains(&new_ifaces, iface->name)) {
3200 iface_destroy(iface);
3204 sset_destroy(&new_ifaces);
3208 port_destroy(struct port *port)
3211 struct bridge *br = port->bridge;
3212 struct iface *iface, *next;
3215 ofproto_bundle_unregister(br->ofproto, port);
3218 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3219 iface_destroy__(iface);
3222 hmap_remove(&br->ports, &port->hmap_node);
3228 static struct port *
3229 port_lookup(const struct bridge *br, const char *name)
3233 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3235 if (!strcmp(port->name, name)) {
3243 enable_lacp(struct port *port, bool *activep)
3245 if (!port->cfg->lacp) {
3246 /* XXX when LACP implementation has been sufficiently tested, enable by
3247 * default and make active on bonded ports. */
3249 } else if (!strcmp(port->cfg->lacp, "off")) {
3251 } else if (!strcmp(port->cfg->lacp, "active")) {
3254 } else if (!strcmp(port->cfg->lacp, "passive")) {
3258 VLOG_WARN("port %s: unknown LACP mode %s",
3259 port->name, port->cfg->lacp);
3264 static struct lacp_settings *
3265 port_configure_lacp(struct port *port, struct lacp_settings *s)
3267 const char *lacp_time, *system_id;
3270 if (!enable_lacp(port, &s->active)) {
3274 s->name = port->name;
3276 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3278 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3279 ETH_ADDR_SCAN_ARGS(s->id))) {
3280 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3281 " address.", port->name, system_id);
3285 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3288 if (eth_addr_is_zero(s->id)) {
3289 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3293 /* Prefer bondable links if unspecified. */
3294 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3296 s->priority = (priority > 0 && priority <= UINT16_MAX
3298 : UINT16_MAX - !list_is_short(&port->ifaces));
3300 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3301 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3303 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3304 "lacp-fallback-ab", false);
3310 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3312 int priority, portid, key;
3314 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3315 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3317 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3319 if (portid <= 0 || portid > UINT16_MAX) {
3320 portid = ofp_to_u16(iface->ofp_port);
3323 if (priority <= 0 || priority > UINT16_MAX) {
3324 priority = UINT16_MAX;
3327 if (key < 0 || key > UINT16_MAX) {
3331 s->name = iface->name;
3333 s->priority = priority;
3338 port_configure_bond(struct port *port, struct bond_settings *s)
3340 const char *detect_s;
3341 struct iface *iface;
3342 int miimon_interval;
3344 s->name = port->name;
3346 if (port->cfg->bond_mode) {
3347 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3348 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3349 port->name, port->cfg->bond_mode,
3350 bond_mode_to_string(s->balance));
3353 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3355 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3356 * active-backup. At some point we should remove this warning. */
3357 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3358 " in previous versions, the default bond_mode was"
3359 " balance-slb", port->name,
3360 bond_mode_to_string(s->balance));
3362 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3363 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3364 "please use another bond type or disable flood_vlans",
3368 miimon_interval = smap_get_int(&port->cfg->other_config,
3369 "bond-miimon-interval", 0);
3370 if (miimon_interval <= 0) {
3371 miimon_interval = 200;
3374 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3375 if (!detect_s || !strcmp(detect_s, "carrier")) {
3376 miimon_interval = 0;
3377 } else if (strcmp(detect_s, "miimon")) {
3378 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3379 "defaulting to carrier", port->name, detect_s);
3380 miimon_interval = 0;
3383 s->up_delay = MAX(0, port->cfg->bond_updelay);
3384 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3385 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3386 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3387 "bond-rebalance-interval", 10000);
3388 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3389 s->rebalance_interval = 1000;
3392 s->fake_iface = port->cfg->bond_fake_iface;
3394 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3395 "lacp-fallback-ab", false);
3397 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3398 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3402 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3403 * instead of obtaining it from the database. */
3405 port_is_synthetic(const struct port *port)
3407 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3410 /* Interface functions. */
3413 iface_is_internal(const struct ovsrec_interface *iface,
3414 const struct ovsrec_bridge *br)
3416 /* The local port and "internal" ports are always "internal". */
3417 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3420 /* Returns the correct network device type for interface 'iface' in bridge
3423 iface_get_type(const struct ovsrec_interface *iface,
3424 const struct ovsrec_bridge *br)
3428 /* The local port always has type "internal". Other ports take
3429 * their type from the database and default to "system" if none is
3431 if (iface_is_internal(iface, br)) {
3434 type = iface->type[0] ? iface->type : "system";
3437 return ofproto_port_open_type(br->datapath_type, type);
3441 iface_destroy__(struct iface *iface)
3444 struct port *port = iface->port;
3445 struct bridge *br = port->bridge;
3447 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3448 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3451 if (iface->ofp_port != OFPP_NONE) {
3452 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3455 list_remove(&iface->port_elem);
3456 hmap_remove(&br->iface_by_name, &iface->name_node);
3458 netdev_close(iface->netdev);
3466 iface_destroy(struct iface *iface)
3469 struct port *port = iface->port;
3471 iface_destroy__(iface);
3472 if (list_is_empty(&port->ifaces)) {
3478 static struct iface *
3479 iface_lookup(const struct bridge *br, const char *name)
3481 struct iface *iface;
3483 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3484 &br->iface_by_name) {
3485 if (!strcmp(iface->name, name)) {
3493 static struct iface *
3494 iface_find(const char *name)
3496 const struct bridge *br;
3498 HMAP_FOR_EACH (br, node, &all_bridges) {
3499 struct iface *iface = iface_lookup(br, name);
3508 static struct iface *
3509 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3511 struct iface *iface;
3513 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3515 if (iface->ofp_port == ofp_port) {
3522 /* Set Ethernet address of 'iface', if one is specified in the configuration
3525 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3527 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3528 struct iface *hw_addr_iface;
3530 if (strcmp(iface->type, "internal")) {
3534 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3536 } else if (port->cfg->fake_bridge) {
3537 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3538 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3543 if (iface->ofp_port == OFPP_LOCAL) {
3544 VLOG_ERR("interface %s: ignoring mac in Interface record "
3545 "(use Bridge record to set local port's mac)",
3547 } else if (eth_addr_is_multicast(mac)) {
3548 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3551 int error = netdev_set_etheraddr(iface->netdev, mac);
3553 VLOG_ERR("interface %s: setting MAC failed (%s)",
3554 iface->name, ovs_strerror(error));
3560 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3562 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3564 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3565 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
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 iface_set_ofport(if_cfg, OFPP_NONE);
3580 ovsrec_interface_set_status(if_cfg, NULL);
3581 ovsrec_interface_set_admin_state(if_cfg, NULL);
3582 ovsrec_interface_set_duplex(if_cfg, NULL);
3583 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3584 ovsrec_interface_set_link_state(if_cfg, NULL);
3585 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3586 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3587 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3588 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3589 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3590 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3591 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3592 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3597 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3599 union ovsdb_atom atom;
3601 atom.integer = target;
3602 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3606 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3608 struct ofpbuf queues_buf;
3610 ofpbuf_init(&queues_buf, 0);
3612 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3613 netdev_set_qos(iface->netdev, NULL, NULL);
3615 const struct ovsdb_datum *queues;
3616 struct netdev_queue_dump dump;
3617 unsigned int queue_id;
3618 struct smap details;
3622 /* Configure top-level Qos for 'iface'. */
3623 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3625 /* Deconfigure queues that were deleted. */
3626 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3628 smap_init(&details);
3629 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3630 if (!queue_ids_include(queues, queue_id)) {
3631 netdev_delete_queue(iface->netdev, queue_id);
3634 smap_destroy(&details);
3636 /* Configure queues for 'iface'. */
3638 for (i = 0; i < qos->n_queues; i++) {
3639 const struct ovsrec_queue *queue = qos->value_queues[i];
3640 unsigned int queue_id = qos->key_queues[i];
3642 if (queue_id == 0) {
3646 if (queue->n_dscp == 1) {
3647 struct ofproto_port_queue *port_queue;
3649 port_queue = ofpbuf_put_uninit(&queues_buf,
3650 sizeof *port_queue);
3651 port_queue->queue = queue_id;
3652 port_queue->dscp = queue->dscp[0];
3655 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3658 struct smap details;
3660 smap_init(&details);
3661 netdev_set_queue(iface->netdev, 0, &details);
3662 smap_destroy(&details);
3666 if (iface->ofp_port != OFPP_NONE) {
3667 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3668 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3670 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3671 port_queues, n_queues);
3674 netdev_set_policing(iface->netdev,
3675 iface->cfg->ingress_policing_rate,
3676 iface->cfg->ingress_policing_burst);
3678 ofpbuf_uninit(&queues_buf);
3682 iface_configure_cfm(struct iface *iface)
3684 const struct ovsrec_interface *cfg = iface->cfg;
3685 const char *opstate_str;
3686 const char *cfm_ccm_vlan;
3687 struct cfm_settings s;
3688 struct smap netdev_args;
3690 if (!cfg->n_cfm_mpid) {
3691 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3695 s.check_tnl_key = false;
3696 smap_init(&netdev_args);
3697 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3698 const char *key = smap_get(&netdev_args, "key");
3699 const char *in_key = smap_get(&netdev_args, "in_key");
3701 s.check_tnl_key = (key && !strcmp(key, "flow"))
3702 || (in_key && !strcmp(in_key, "flow"));
3704 smap_destroy(&netdev_args);
3706 s.mpid = *cfg->cfm_mpid;
3707 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3708 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3709 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3711 if (s.interval <= 0) {
3715 if (!cfm_ccm_vlan) {
3717 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3718 s.ccm_vlan = CFM_RANDOM_VLAN;
3720 s.ccm_vlan = atoi(cfm_ccm_vlan);
3721 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3726 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3728 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3730 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3731 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3733 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3736 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3737 * instead of obtaining it from the database. */
3739 iface_is_synthetic(const struct iface *iface)
3741 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3745 iface_validate_ofport__(size_t n, int64_t *ofport)
3747 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3748 ? u16_to_ofp(*ofport)
3753 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3755 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3759 iface_pick_ofport(const struct ovsrec_interface *cfg)
3761 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3762 return (requested_ofport != OFPP_NONE
3764 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3767 /* Port mirroring. */
3769 static struct mirror *
3770 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3774 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3775 if (uuid_equals(uuid, &m->uuid)) {
3783 bridge_configure_mirrors(struct bridge *br)
3785 const struct ovsdb_datum *mc;
3786 unsigned long *flood_vlans;
3787 struct mirror *m, *next;
3790 /* Get rid of deleted mirrors. */
3791 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3792 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3793 union ovsdb_atom atom;
3795 atom.uuid = m->uuid;
3796 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3801 /* Add new mirrors and reconfigure existing ones. */
3802 for (i = 0; i < br->cfg->n_mirrors; i++) {
3803 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3804 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3806 m = mirror_create(br, cfg);
3809 if (!mirror_configure(m)) {
3814 /* Update flooded vlans (for RSPAN). */
3815 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3816 br->cfg->n_flood_vlans);
3817 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3818 bitmap_free(flood_vlans);
3821 static struct mirror *
3822 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3826 m = xzalloc(sizeof *m);
3827 m->uuid = cfg->header_.uuid;
3828 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3830 m->name = xstrdup(cfg->name);
3836 mirror_destroy(struct mirror *m)
3839 struct bridge *br = m->bridge;
3842 ofproto_mirror_unregister(br->ofproto, m);
3845 hmap_remove(&br->mirrors, &m->hmap_node);
3852 mirror_collect_ports(struct mirror *m,
3853 struct ovsrec_port **in_ports, int n_in_ports,
3854 void ***out_portsp, size_t *n_out_portsp)
3856 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3857 size_t n_out_ports = 0;
3860 for (i = 0; i < n_in_ports; i++) {
3861 const char *name = in_ports[i]->name;
3862 struct port *port = port_lookup(m->bridge, name);
3864 out_ports[n_out_ports++] = port;
3866 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3867 "port %s", m->bridge->name, m->name, name);
3870 *out_portsp = out_ports;
3871 *n_out_portsp = n_out_ports;
3875 mirror_configure(struct mirror *m)
3877 const struct ovsrec_mirror *cfg = m->cfg;
3878 struct ofproto_mirror_settings s;
3881 if (strcmp(cfg->name, m->name)) {
3883 m->name = xstrdup(cfg->name);
3887 /* Get output port or VLAN. */
3888 if (cfg->output_port) {
3889 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3890 if (!s.out_bundle) {
3891 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3892 m->bridge->name, m->name);
3895 s.out_vlan = UINT16_MAX;
3897 if (cfg->output_vlan) {
3898 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3899 "output vlan; ignoring output vlan",
3900 m->bridge->name, m->name);
3902 } else if (cfg->output_vlan) {
3903 /* The database should prevent invalid VLAN values. */
3904 s.out_bundle = NULL;
3905 s.out_vlan = *cfg->output_vlan;
3907 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3908 m->bridge->name, m->name);
3912 /* Get port selection. */
3913 if (cfg->select_all) {
3914 size_t n_ports = hmap_count(&m->bridge->ports);
3915 void **ports = xmalloc(n_ports * sizeof *ports);
3920 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3930 /* Get ports, dropping ports that don't exist.
3931 * The IDL ensures that there are no duplicates. */
3932 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3933 &s.srcs, &s.n_srcs);
3934 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3935 &s.dsts, &s.n_dsts);
3938 /* Get VLAN selection. */
3939 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3942 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3945 if (s.srcs != s.dsts) {
3954 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3956 * This is deprecated. It is only for compatibility with broken device drivers
3957 * in old versions of Linux that do not properly support VLANs when VLAN
3958 * devices are not used. When broken device drivers are no longer in
3959 * widespread use, we will delete these interfaces. */
3961 static struct ovsrec_port **recs;
3962 static size_t n_recs, allocated_recs;
3964 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3965 * splinters are reconfigured. */
3967 register_rec(struct ovsrec_port *rec)
3969 if (n_recs >= allocated_recs) {
3970 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3972 recs[n_recs++] = rec;
3975 /* Frees all of the ports registered with register_reg(). */
3977 free_registered_recs(void)
3981 for (i = 0; i < n_recs; i++) {
3982 struct ovsrec_port *port = recs[i];
3985 for (j = 0; j < port->n_interfaces; j++) {
3986 struct ovsrec_interface *iface = port->interfaces[j];
3991 smap_destroy(&port->other_config);
3992 free(port->interfaces);
4000 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4003 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4005 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4009 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4012 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4013 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4014 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4017 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4018 * use, returns NULL.
4020 * Updates 'vlan_splinters_enabled_anywhere'. */
4021 static unsigned long int *
4022 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4024 unsigned long int *splinter_vlans;
4025 struct sset splinter_ifaces;
4026 const char *real_dev_name;
4027 struct shash *real_devs;
4028 struct shash_node *node;
4032 /* Free space allocated for synthesized ports and interfaces, since we're
4033 * in the process of reconstructing all of them. */
4034 free_registered_recs();
4036 splinter_vlans = bitmap_allocate(4096);
4037 sset_init(&splinter_ifaces);
4038 vlan_splinters_enabled_anywhere = false;
4039 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4040 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4043 for (j = 0; j < br_cfg->n_ports; j++) {
4044 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4047 for (k = 0; k < port_cfg->n_interfaces; k++) {
4048 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4050 if (vlan_splinters_is_enabled(iface_cfg)) {
4051 vlan_splinters_enabled_anywhere = true;
4052 sset_add(&splinter_ifaces, iface_cfg->name);
4053 vlan_bitmap_from_array__(port_cfg->trunks,
4059 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4060 bitmap_set1(splinter_vlans, *port_cfg->tag);
4065 if (!vlan_splinters_enabled_anywhere) {
4066 free(splinter_vlans);
4067 sset_destroy(&splinter_ifaces);
4071 HMAP_FOR_EACH (br, node, &all_bridges) {
4073 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4077 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4078 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4079 * device to be created for it. */
4080 bitmap_set0(splinter_vlans, 0);
4081 bitmap_set0(splinter_vlans, 4095);
4083 /* Delete all VLAN devices that we don't need. */
4085 real_devs = vlandev_get_real_devs();
4086 SHASH_FOR_EACH (node, real_devs) {
4087 const struct vlan_real_dev *real_dev = node->data;
4088 const struct vlan_dev *vlan_dev;
4089 bool real_dev_has_splinters;
4091 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4093 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4094 if (!real_dev_has_splinters
4095 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4096 struct netdev *netdev;
4098 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4099 if (!netdev_get_in4(netdev, NULL, NULL) ||
4100 !netdev_get_in6(netdev, NULL)) {
4101 /* It has an IP address configured, so we don't own
4102 * it. Don't delete it. */
4104 vlandev_del(vlan_dev->name);
4106 netdev_close(netdev);
4113 /* Add all VLAN devices that we need. */
4114 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4117 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4118 if (!vlandev_get_name(real_dev_name, vid)) {
4119 vlandev_add(real_dev_name, vid);
4126 sset_destroy(&splinter_ifaces);
4128 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4129 free(splinter_vlans);
4132 return splinter_vlans;
4135 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4138 configure_splinter_port(struct port *port)
4140 struct ofproto *ofproto = port->bridge->ofproto;
4141 ofp_port_t realdev_ofp_port;
4142 const char *realdev_name;
4143 struct iface *vlandev, *realdev;
4145 ofproto_bundle_unregister(port->bridge->ofproto, port);
4147 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4150 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4151 realdev = iface_lookup(port->bridge, realdev_name);
4152 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4154 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4158 static struct ovsrec_port *
4159 synthesize_splinter_port(const char *real_dev_name,
4160 const char *vlan_dev_name, int vid)
4162 struct ovsrec_interface *iface;
4163 struct ovsrec_port *port;
4165 iface = xmalloc(sizeof *iface);
4166 ovsrec_interface_init(iface);
4167 iface->name = xstrdup(vlan_dev_name);
4168 iface->type = "system";
4170 port = xmalloc(sizeof *port);
4171 ovsrec_port_init(port);
4172 port->interfaces = xmemdup(&iface, sizeof iface);
4173 port->n_interfaces = 1;
4174 port->name = xstrdup(vlan_dev_name);
4175 port->vlan_mode = "splinter";
4176 port->tag = xmalloc(sizeof *port->tag);
4179 smap_add(&port->other_config, "realdev", real_dev_name);
4185 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4186 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4187 * 1-bit in the 'splinter_vlans' bitmap. */
4189 add_vlan_splinter_ports(struct bridge *br,
4190 const unsigned long int *splinter_vlans,
4191 struct shash *ports)
4195 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4196 * we're modifying 'ports'. */
4197 for (i = 0; i < br->cfg->n_ports; i++) {
4198 const char *name = br->cfg->ports[i]->name;
4199 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4202 for (j = 0; j < port_cfg->n_interfaces; j++) {
4203 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4205 if (vlan_splinters_is_enabled(iface_cfg)) {
4206 const char *real_dev_name;
4209 real_dev_name = iface_cfg->name;
4210 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4211 const char *vlan_dev_name;
4213 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4215 && !shash_find(ports, vlan_dev_name)) {
4216 shash_add(ports, vlan_dev_name,
4217 synthesize_splinter_port(
4218 real_dev_name, vlan_dev_name, vid));
4227 mirror_refresh_stats(struct mirror *m)
4229 struct ofproto *ofproto = m->bridge->ofproto;
4230 uint64_t tx_packets, tx_bytes;
4233 size_t stat_cnt = 0;
4235 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4236 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4240 if (tx_packets != UINT64_MAX) {
4241 keys[stat_cnt] = "tx_packets";
4242 values[stat_cnt] = tx_packets;
4245 if (tx_bytes != UINT64_MAX) {
4246 keys[stat_cnt] = "tx_bytes";
4247 values[stat_cnt] = tx_bytes;
4251 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);