1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 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.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
55 #include "socket-util.h"
57 #include "stream-ssl.h"
59 #include "system-stats.h"
64 #include "lib/vswitch-idl.h"
65 #include "xenserver.h"
66 #include "openvswitch/vlog.h"
67 #include "sflow_api.h"
68 #include "vlan-bitmap.h"
71 VLOG_DEFINE_THIS_MODULE(bridge);
73 COVERAGE_DEFINE(bridge_reconfigure);
76 /* These members are always valid.
78 * They are immutable: they never change between iface_create() and
80 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
85 struct netdev *netdev; /* Network device. */
86 ofp_port_t ofp_port; /* OpenFlow port number. */
89 /* These members are valid only within bridge_reconfigure(). */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct ovs_list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 /* Port mirroring. */
131 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
134 struct hmap mappings; /* "struct" indexed by UUID */
136 /* Used during reconfiguration. */
137 struct shash wanted_ports;
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
146 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
147 struct bridge *bridge;
153 /* All bridges, indexed by name. */
154 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
156 /* OVSDB IDL used to obtain configuration. */
157 static struct ovsdb_idl *idl;
159 /* We want to complete daemonization, fully detaching from our parent process,
160 * only after we have completed our initial configuration, committed our state
161 * to the database, and received confirmation back from the database server
162 * that it applied the commit. This allows our parent process to know that,
163 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
164 * to have already been filled in. (It is only "very likely" rather than
165 * certain because there is always a slim possibility that the transaction will
166 * fail or that some other client has added new bridges, ports, etc. while
167 * ovs-vswitchd was configuring using an old configuration.)
169 * We only need to do this once for our initial configuration at startup, so
170 * 'initial_config_done' tracks whether we've already done it. While we are
171 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
172 * itself and is otherwise NULL. */
173 static bool initial_config_done;
174 static struct ovsdb_idl_txn *daemonize_txn;
176 /* Most recently processed IDL sequence number. */
177 static unsigned int idl_seqno;
179 /* Track changes to port connectivity. */
180 static uint64_t connectivity_seqno = LLONG_MIN;
182 /* Status update to database.
184 * Some information in the database must be kept as up-to-date as possible to
185 * allow controllers to respond rapidly to network outages. Those status are
186 * updated via the 'status_txn'.
188 * We use the global connectivity sequence number to detect the status change.
189 * Also, to prevent the status update from sending too much to the database,
190 * we check the return status of each update transaction and do not start new
191 * update if the previous transaction status is 'TXN_INCOMPLETE'.
193 * 'statux_txn' is NULL if there is no ongoing status update.
195 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
196 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
197 * which will cause the main thread wake up soon and retry the status update.
199 static struct ovsdb_idl_txn *status_txn;
200 static bool status_txn_try_again;
202 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
203 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
204 #define STATUS_CHECK_AGAIN_MSEC 100
206 /* Statistics update to database. */
207 static struct ovsdb_idl_txn *stats_txn;
209 /* Each time this timer expires, the bridge fetches interface and mirror
210 * statistics and pushes them into the database. */
211 static int stats_timer_interval;
212 static long long int stats_timer = LLONG_MIN;
214 /* Each time this timer expires, the bridge fetches the list of port/VLAN
215 * membership that has been modified by the AA.
217 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
218 static long long int aa_refresh_timer = LLONG_MIN;
220 static void add_del_bridges(const struct ovsrec_open_vswitch *);
221 static void bridge_run__(void);
222 static void bridge_create(const struct ovsrec_bridge *);
223 static void bridge_destroy(struct bridge *);
224 static struct bridge *bridge_lookup(const char *name);
225 static unixctl_cb_func bridge_unixctl_dump_flows;
226 static unixctl_cb_func bridge_unixctl_reconnect;
227 static size_t bridge_get_controllers(const struct bridge *br,
228 struct ovsrec_controller ***controllersp);
229 static void bridge_collect_wanted_ports(struct bridge *,
230 const unsigned long *splinter_vlans,
231 struct shash *wanted_ports);
232 static void bridge_delete_ofprotos(void);
233 static void bridge_delete_or_reconfigure_ports(struct bridge *);
234 static void bridge_del_ports(struct bridge *,
235 const struct shash *wanted_ports);
236 static void bridge_add_ports(struct bridge *,
237 const struct shash *wanted_ports);
239 static void bridge_configure_datapath_id(struct bridge *);
240 static void bridge_configure_netflow(struct bridge *);
241 static void bridge_configure_forward_bpdu(struct bridge *);
242 static void bridge_configure_mac_table(struct bridge *);
243 static void bridge_configure_mcast_snooping(struct bridge *);
244 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
245 static void bridge_configure_ipfix(struct bridge *);
246 static void bridge_configure_spanning_tree(struct bridge *);
247 static void bridge_configure_tables(struct bridge *);
248 static void bridge_configure_dp_desc(struct bridge *);
249 static void bridge_configure_aa(struct bridge *);
250 static void bridge_aa_refresh_queued(struct bridge *);
251 static bool bridge_aa_need_refresh(struct bridge *);
252 static void bridge_configure_remotes(struct bridge *,
253 const struct sockaddr_in *managers,
255 static void bridge_pick_local_hw_addr(struct bridge *,
256 uint8_t ea[ETH_ADDR_LEN],
257 struct iface **hw_addr_iface);
258 static uint64_t bridge_pick_datapath_id(struct bridge *,
259 const uint8_t bridge_ea[ETH_ADDR_LEN],
260 struct iface *hw_addr_iface);
261 static uint64_t dpid_from_hash(const void *, size_t nbytes);
262 static bool bridge_has_bond_fake_iface(const struct bridge *,
264 static bool port_is_bond_fake_iface(const struct port *);
266 static unixctl_cb_func qos_unixctl_show;
268 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
269 static void port_del_ifaces(struct port *);
270 static void port_destroy(struct port *);
271 static struct port *port_lookup(const struct bridge *, const char *name);
272 static void port_configure(struct port *);
273 static struct lacp_settings *port_configure_lacp(struct port *,
274 struct lacp_settings *);
275 static void port_configure_bond(struct port *, struct bond_settings *);
276 static bool port_is_synthetic(const struct port *);
278 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
279 static void run_system_stats(void);
281 static void bridge_configure_mirrors(struct bridge *);
282 static struct mirror *mirror_create(struct bridge *,
283 const struct ovsrec_mirror *);
284 static void mirror_destroy(struct mirror *);
285 static bool mirror_configure(struct mirror *);
286 static void mirror_refresh_stats(struct mirror *);
288 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
289 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
290 const struct ovsrec_port *);
291 static bool iface_is_internal(const struct ovsrec_interface *iface,
292 const struct ovsrec_bridge *br);
293 static const char *iface_get_type(const struct ovsrec_interface *,
294 const struct ovsrec_bridge *);
295 static void iface_destroy(struct iface *);
296 static void iface_destroy__(struct iface *);
297 static struct iface *iface_lookup(const struct bridge *, const char *name);
298 static struct iface *iface_find(const char *name);
299 static struct iface *iface_from_ofp_port(const struct bridge *,
300 ofp_port_t ofp_port);
301 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
302 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
303 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
304 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
305 static void iface_configure_cfm(struct iface *);
306 static void iface_refresh_cfm_stats(struct iface *);
307 static void iface_refresh_stats(struct iface *);
308 static void iface_refresh_netdev_status(struct iface *);
309 static void iface_refresh_ofproto_status(struct iface *);
310 static bool iface_is_synthetic(const struct iface *);
311 static ofp_port_t iface_get_requested_ofp_port(
312 const struct ovsrec_interface *);
313 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
316 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
318 * This is deprecated. It is only for compatibility with broken device drivers
319 * in old versions of Linux that do not properly support VLANs when VLAN
320 * devices are not used. When broken device drivers are no longer in
321 * widespread use, we will delete these interfaces. */
323 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
324 static bool vlan_splinters_enabled_anywhere;
326 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
327 static unsigned long int *collect_splinter_vlans(
328 const struct ovsrec_open_vswitch *);
329 static void configure_splinter_port(struct port *);
330 static void add_vlan_splinter_ports(struct bridge *,
331 const unsigned long int *splinter_vlans,
332 struct shash *ports);
334 static void discover_types(const struct ovsrec_open_vswitch *cfg);
337 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
339 struct shash iface_hints;
340 static bool initialized = false;
347 shash_init(&iface_hints);
350 for (i = 0; i < cfg->n_bridges; i++) {
351 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
354 for (j = 0; j < br_cfg->n_ports; j++) {
355 struct ovsrec_port *port_cfg = br_cfg->ports[j];
358 for (k = 0; k < port_cfg->n_interfaces; k++) {
359 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
360 struct iface_hint *iface_hint;
362 iface_hint = xmalloc(sizeof *iface_hint);
363 iface_hint->br_name = br_cfg->name;
364 iface_hint->br_type = br_cfg->datapath_type;
365 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
367 shash_add(&iface_hints, if_cfg->name, iface_hint);
373 ofproto_init(&iface_hints);
375 shash_destroy_free_data(&iface_hints);
379 /* Public functions. */
381 /* Initializes the bridge module, configuring it to obtain its configuration
382 * from an OVSDB server accessed over 'remote', which should be a string in a
383 * form acceptable to ovsdb_idl_create(). */
385 bridge_init(const char *remote)
387 /* Create connection to database. */
388 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
389 idl_seqno = ovsdb_idl_get_seqno(idl);
390 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
391 ovsdb_idl_verify_write_only(idl);
393 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
394 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
395 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
396 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
397 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
399 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
400 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
401 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
403 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
404 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
405 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
406 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
407 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
408 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
409 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
411 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
412 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
413 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
414 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
415 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
416 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
417 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
418 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
419 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
420 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
421 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
422 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
423 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
424 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
425 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
426 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
427 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
428 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
439 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
441 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
442 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
443 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
444 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
446 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
448 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
450 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
451 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
453 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
454 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
455 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
456 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
458 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
459 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
460 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
461 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
462 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
464 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
466 /* Register unixctl commands. */
467 unixctl_command_register("qos/show", "interface", 1, 1,
468 qos_unixctl_show, NULL);
469 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
470 bridge_unixctl_dump_flows, NULL);
471 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
472 bridge_unixctl_reconnect, NULL);
486 struct bridge *br, *next_br;
488 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
491 ovsdb_idl_destroy(idl);
494 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
495 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
496 * responsible for freeing '*managersp' (with free()).
498 * You may be asking yourself "why does ovs-vswitchd care?", because
499 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
500 * should not be and in fact is not directly involved in that. But
501 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
502 * it has to tell in-band control where the managers are to enable that.
503 * (Thus, only managers connected in-band and with non-loopback addresses
507 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
508 struct sockaddr_in **managersp, size_t *n_managersp)
510 struct sockaddr_in *managers = NULL;
511 size_t n_managers = 0;
515 /* Collect all of the potential targets from the "targets" columns of the
516 * rows pointed to by "manager_options", excluding any that are
519 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
520 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
522 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
523 sset_find_and_delete(&targets, m->target);
525 sset_add(&targets, m->target);
529 /* Now extract the targets' IP addresses. */
530 if (!sset_is_empty(&targets)) {
533 managers = xmalloc(sset_count(&targets) * sizeof *managers);
534 SSET_FOR_EACH (target, &targets) {
536 struct sockaddr_storage ss;
537 struct sockaddr_in in;
540 /* Ignore loopback. */
541 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
543 && sa.ss.ss_family == AF_INET
544 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
545 managers[n_managers++] = sa.in;
549 sset_destroy(&targets);
551 *managersp = managers;
552 *n_managersp = n_managers;
556 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
558 unsigned long int *splinter_vlans;
559 struct sockaddr_in *managers;
560 struct bridge *br, *next;
561 int sflow_bridge_number;
564 COVERAGE_INC(bridge_reconfigure);
566 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
567 OFPROTO_FLOW_LIMIT_DEFAULT));
568 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
569 OFPROTO_MAX_IDLE_DEFAULT));
570 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
572 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
575 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
576 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
578 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
579 * to 'ovs_cfg', with only very minimal configuration otherwise.
581 * This is mostly an update to bridge data structures. Nothing is pushed
582 * down to ofproto or lower layers. */
583 add_del_bridges(ovs_cfg);
584 splinter_vlans = collect_splinter_vlans(ovs_cfg);
585 HMAP_FOR_EACH (br, node, &all_bridges) {
586 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
587 bridge_del_ports(br, &br->wanted_ports);
589 free(splinter_vlans);
591 /* Start pushing configuration changes down to the ofproto layer:
593 * - Delete ofprotos that are no longer configured.
595 * - Delete ports that are no longer configured.
597 * - Reconfigure existing ports to their desired configurations, or
598 * delete them if not possible.
600 * We have to do all the deletions before we can do any additions, because
601 * the ports to be added might require resources that will be freed up by
602 * deletions (they might especially overlap in name). */
603 bridge_delete_ofprotos();
604 HMAP_FOR_EACH (br, node, &all_bridges) {
606 bridge_delete_or_reconfigure_ports(br);
610 /* Finish pushing configuration changes to the ofproto layer:
612 * - Create ofprotos that are missing.
614 * - Add ports that are missing. */
615 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
619 error = ofproto_create(br->name, br->type, &br->ofproto);
621 VLOG_ERR("failed to create bridge %s: %s", br->name,
622 ovs_strerror(error));
623 shash_destroy(&br->wanted_ports);
626 /* Trigger storing datapath version. */
627 seq_change(connectivity_seq_get());
631 HMAP_FOR_EACH (br, node, &all_bridges) {
632 bridge_add_ports(br, &br->wanted_ports);
633 shash_destroy(&br->wanted_ports);
636 reconfigure_system_stats(ovs_cfg);
638 /* Complete the configuration. */
639 sflow_bridge_number = 0;
640 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
641 HMAP_FOR_EACH (br, node, &all_bridges) {
644 /* We need the datapath ID early to allow LACP ports to use it as the
645 * default system ID. */
646 bridge_configure_datapath_id(br);
648 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
651 port_configure(port);
653 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
654 iface_set_ofport(iface->cfg, iface->ofp_port);
655 /* Clear eventual previous errors */
656 ovsrec_interface_set_error(iface->cfg, NULL);
657 iface_configure_cfm(iface);
658 iface_configure_qos(iface, port->cfg->qos);
659 iface_set_mac(br, port, iface);
660 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
662 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
666 bridge_configure_mirrors(br);
667 bridge_configure_forward_bpdu(br);
668 bridge_configure_mac_table(br);
669 bridge_configure_mcast_snooping(br);
670 bridge_configure_remotes(br, managers, n_managers);
671 bridge_configure_netflow(br);
672 bridge_configure_sflow(br, &sflow_bridge_number);
673 bridge_configure_ipfix(br);
674 bridge_configure_spanning_tree(br);
675 bridge_configure_tables(br);
676 bridge_configure_dp_desc(br);
677 bridge_configure_aa(br);
681 /* The ofproto-dpif provider does some final reconfiguration in its
682 * ->type_run() function. We have to call it before notifying the database
683 * client that reconfiguration is complete, otherwise there is a very
684 * narrow race window in which e.g. ofproto/trace will not recognize the
685 * new configuration (sometimes this causes unit test failures). */
689 /* Delete ofprotos which aren't configured or have the wrong type. Create
690 * ofprotos which don't exist but need to. */
692 bridge_delete_ofprotos(void)
699 /* Delete ofprotos with no bridge or with the wrong type. */
702 ofproto_enumerate_types(&types);
703 SSET_FOR_EACH (type, &types) {
706 ofproto_enumerate_names(type, &names);
707 SSET_FOR_EACH (name, &names) {
708 br = bridge_lookup(name);
709 if (!br || strcmp(type, br->type)) {
710 ofproto_delete(name, type);
714 sset_destroy(&names);
715 sset_destroy(&types);
719 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
721 if (*n >= *allocated) {
722 ports = x2nrealloc(ports, allocated, sizeof *ports);
724 ports[(*n)++] = port;
729 bridge_delete_or_reconfigure_ports(struct bridge *br)
731 struct ofproto_port ofproto_port;
732 struct ofproto_port_dump dump;
734 struct sset ofproto_ports;
735 struct port *port, *port_next;
737 /* List of "ofp_port"s to delete. We make a list instead of deleting them
738 * right away because ofproto implementations aren't necessarily able to
739 * iterate through a changing list of ports in an entirely robust way. */
746 sset_init(&ofproto_ports);
748 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
749 * that are not configured in the database. (This commonly happens when
750 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
752 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
753 * that have the wrong OpenFlow port number (and arrange to add them back
754 * with the correct OpenFlow port number). */
755 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
756 ofp_port_t requested_ofp_port;
759 sset_add(&ofproto_ports, ofproto_port.name);
761 iface = iface_lookup(br, ofproto_port.name);
763 /* No such iface is configured, so we should delete this
766 * As a corner case exception, keep the port if it's a bond fake
768 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
769 && !strcmp(ofproto_port.type, "internal")) {
775 if (strcmp(ofproto_port.type, iface->type)
776 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
777 /* The interface is the wrong type or can't be configured.
782 /* If the requested OpenFlow port for 'iface' changed, and it's not
783 * already the correct port, then we might want to temporarily delete
784 * this interface, so we can add it back again with the new OpenFlow
786 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
787 if (iface->ofp_port != OFPP_LOCAL &&
788 requested_ofp_port != OFPP_NONE &&
789 requested_ofp_port != iface->ofp_port) {
790 ofp_port_t victim_request;
791 struct iface *victim;
793 /* Check for an existing OpenFlow port currently occupying
794 * 'iface''s requested port number. If there isn't one, then
795 * delete this port. Otherwise we need to consider further. */
796 victim = iface_from_ofp_port(br, requested_ofp_port);
801 /* 'victim' is a port currently using 'iface''s requested port
802 * number. Unless 'victim' specifically requested that port
803 * number, too, then we can delete both 'iface' and 'victim'
804 * temporarily. (We'll add both of them back again later with new
805 * OpenFlow port numbers.)
807 * If 'victim' did request port number 'requested_ofp_port', just
808 * like 'iface', then that's a configuration inconsistency that we
809 * can't resolve. We might as well let it keep its current port
811 victim_request = iface_get_requested_ofp_port(victim->cfg);
812 if (victim_request != requested_ofp_port) {
813 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
814 iface_destroy(victim);
823 iface_destroy(iface);
824 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
826 for (i = 0; i < n; i++) {
827 ofproto_port_del(br->ofproto, del[i]);
831 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
832 * that we didn't see when we iterated through the datapath, i.e. ports
833 * that disappeared underneath use. This is an unusual situation, but it
834 * can happen in some cases:
836 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
837 * idea but could happen).
839 * - The port represented a device that disappeared, e.g. a tuntap
840 * device destroyed via "tunctl -d", a physical Ethernet device
841 * whose module was just unloaded via "rmmod", or a virtual NIC for a
842 * VM whose VM was just terminated. */
843 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
844 struct iface *iface, *iface_next;
846 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
847 if (!sset_contains(&ofproto_ports, iface->name)) {
848 iface_destroy__(iface);
852 if (list_is_empty(&port->ifaces)) {
856 sset_destroy(&ofproto_ports);
860 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
861 bool with_requested_port)
863 struct shash_node *port_node;
865 SHASH_FOR_EACH (port_node, wanted_ports) {
866 const struct ovsrec_port *port_cfg = port_node->data;
869 for (i = 0; i < port_cfg->n_interfaces; i++) {
870 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
871 ofp_port_t requested_ofp_port;
873 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
874 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
875 struct iface *iface = iface_lookup(br, iface_cfg->name);
878 iface_create(br, iface_cfg, port_cfg);
886 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
888 /* First add interfaces that request a particular port number. */
889 bridge_add_ports__(br, wanted_ports, true);
891 /* Then add interfaces that want automatic port number assignment.
892 * We add these afterward to avoid accidentally taking a specifically
893 * requested port number. */
894 bridge_add_ports__(br, wanted_ports, false);
898 port_configure(struct port *port)
900 const struct ovsrec_port *cfg = port->cfg;
901 struct bond_settings bond_settings;
902 struct lacp_settings lacp_settings;
903 struct ofproto_bundle_settings s;
906 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
907 configure_splinter_port(port);
916 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
917 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
918 s.slaves[s.n_slaves++] = iface->ofp_port;
923 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
927 /* Get VLAN trunks. */
930 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
934 if (cfg->vlan_mode) {
935 if (!strcmp(cfg->vlan_mode, "access")) {
936 s.vlan_mode = PORT_VLAN_ACCESS;
937 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
938 s.vlan_mode = PORT_VLAN_TRUNK;
939 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
940 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
941 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
942 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
944 /* This "can't happen" because ovsdb-server should prevent it. */
945 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
946 "back to trunk mode", port->name, cfg->vlan_mode);
947 s.vlan_mode = PORT_VLAN_TRUNK;
951 s.vlan_mode = PORT_VLAN_ACCESS;
953 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
957 s.vlan_mode = PORT_VLAN_TRUNK;
960 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
963 /* Get LACP settings. */
964 s.lacp = port_configure_lacp(port, &lacp_settings);
968 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
969 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
970 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
973 s.lacp_slaves = NULL;
976 /* Get bond settings. */
977 if (s.n_slaves > 1) {
978 s.bond = &bond_settings;
979 port_configure_bond(port, &bond_settings);
982 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
983 netdev_set_miimon_interval(iface->netdev, 0);
988 ofproto_bundle_register(port->bridge->ofproto, port, &s);
996 /* Pick local port hardware address and datapath ID for 'br'. */
998 bridge_configure_datapath_id(struct bridge *br)
1000 uint8_t ea[ETH_ADDR_LEN];
1002 struct iface *local_iface;
1003 struct iface *hw_addr_iface;
1006 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1007 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1009 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1011 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1012 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1013 "Ethernet address: %s",
1014 br->name, ovs_strerror(error));
1017 memcpy(br->ea, ea, ETH_ADDR_LEN);
1019 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1020 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1021 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1022 ofproto_set_datapath_id(br->ofproto, dpid);
1025 dpid_string = xasprintf("%016"PRIx64, dpid);
1026 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1030 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1033 bridge_get_allowed_versions(struct bridge *br)
1035 if (!br->cfg->n_protocols)
1038 return ofputil_versions_from_strings(br->cfg->protocols,
1039 br->cfg->n_protocols);
1042 /* Set NetFlow configuration on 'br'. */
1044 bridge_configure_netflow(struct bridge *br)
1046 struct ovsrec_netflow *cfg = br->cfg->netflow;
1047 struct netflow_options opts;
1050 ofproto_set_netflow(br->ofproto, NULL);
1054 memset(&opts, 0, sizeof opts);
1056 /* Get default NetFlow configuration from datapath.
1057 * Apply overrides from 'cfg'. */
1058 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1059 if (cfg->engine_type) {
1060 opts.engine_type = *cfg->engine_type;
1062 if (cfg->engine_id) {
1063 opts.engine_id = *cfg->engine_id;
1066 /* Configure active timeout interval. */
1067 opts.active_timeout = cfg->active_timeout;
1068 if (!opts.active_timeout) {
1069 opts.active_timeout = -1;
1070 } else if (opts.active_timeout < 0) {
1071 VLOG_WARN("bridge %s: active timeout interval set to negative "
1072 "value, using default instead (%d seconds)", br->name,
1073 NF_ACTIVE_TIMEOUT_DEFAULT);
1074 opts.active_timeout = -1;
1077 /* Add engine ID to interface number to disambiguate bridgs? */
1078 opts.add_id_to_iface = cfg->add_id_to_interface;
1079 if (opts.add_id_to_iface) {
1080 if (opts.engine_id > 0x7f) {
1081 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1082 "another vswitch, choose an engine id less than 128",
1085 if (hmap_count(&br->ports) > 508) {
1086 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1087 "another port when more than 508 ports are used",
1093 sset_init(&opts.collectors);
1094 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1097 if (ofproto_set_netflow(br->ofproto, &opts)) {
1098 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1100 sset_destroy(&opts.collectors);
1103 /* Set sFlow configuration on 'br'. */
1105 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1107 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1108 struct ovsrec_controller **controllers;
1109 struct ofproto_sflow_options oso;
1110 size_t n_controllers;
1114 ofproto_set_sflow(br->ofproto, NULL);
1118 memset(&oso, 0, sizeof oso);
1120 sset_init(&oso.targets);
1121 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1123 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1124 if (cfg->sampling) {
1125 oso.sampling_rate = *cfg->sampling;
1128 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1130 oso.polling_interval = *cfg->polling;
1133 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1135 oso.header_len = *cfg->header;
1138 oso.sub_id = (*sflow_bridge_number)++;
1139 oso.agent_device = cfg->agent;
1141 oso.control_ip = NULL;
1142 n_controllers = bridge_get_controllers(br, &controllers);
1143 for (i = 0; i < n_controllers; i++) {
1144 if (controllers[i]->local_ip) {
1145 oso.control_ip = controllers[i]->local_ip;
1149 ofproto_set_sflow(br->ofproto, &oso);
1151 sset_destroy(&oso.targets);
1154 /* Returns whether a IPFIX row is valid. */
1156 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1158 return ipfix && ipfix->n_targets > 0;
1161 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1163 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1164 const struct bridge *br)
1166 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1169 /* Set IPFIX configuration on 'br'. */
1171 bridge_configure_ipfix(struct bridge *br)
1173 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1174 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1175 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1176 struct ofproto_ipfix_bridge_exporter_options be_opts;
1177 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1178 size_t n_fe_opts = 0;
1180 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1181 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1186 if (!valid_be_cfg && n_fe_opts == 0) {
1187 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1192 memset(&be_opts, 0, sizeof be_opts);
1194 sset_init(&be_opts.targets);
1195 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1197 if (be_cfg->sampling) {
1198 be_opts.sampling_rate = *be_cfg->sampling;
1200 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1202 if (be_cfg->obs_domain_id) {
1203 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1205 if (be_cfg->obs_point_id) {
1206 be_opts.obs_point_id = *be_cfg->obs_point_id;
1208 if (be_cfg->cache_active_timeout) {
1209 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1211 if (be_cfg->cache_max_flows) {
1212 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1215 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1216 "enable-tunnel-sampling", true);
1218 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1219 "enable-input-sampling", false);
1221 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1222 "enable-output-sampling", false);
1225 if (n_fe_opts > 0) {
1226 struct ofproto_ipfix_flow_exporter_options *opts;
1227 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1229 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1230 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1231 opts->collector_set_id = fe_cfg->id;
1232 sset_init(&opts->targets);
1233 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1234 fe_cfg->ipfix->n_targets);
1235 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1236 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1237 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1238 ? *fe_cfg->ipfix->cache_max_flows : 0;
1244 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1248 sset_destroy(&be_opts.targets);
1251 if (n_fe_opts > 0) {
1252 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1254 for (i = 0; i < n_fe_opts; i++) {
1255 sset_destroy(&opts->targets);
1263 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1264 struct ofproto_port_stp_settings *port_s,
1265 int *port_num_counter, unsigned long *port_num_bitmap)
1267 const char *config_str;
1268 struct iface *iface;
1270 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1271 port_s->enable = false;
1274 port_s->enable = true;
1277 /* STP over bonds is not supported. */
1278 if (!list_is_singleton(&port->ifaces)) {
1279 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1281 port_s->enable = false;
1285 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1287 /* Internal ports shouldn't participate in spanning tree, so
1289 if (!strcmp(iface->type, "internal")) {
1290 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1291 port_s->enable = false;
1295 /* STP on mirror output ports is not supported. */
1296 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1297 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1298 port_s->enable = false;
1302 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1304 unsigned long int port_num = strtoul(config_str, NULL, 0);
1305 int port_idx = port_num - 1;
1307 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1308 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1309 port_s->enable = false;
1313 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1314 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1315 port->name, port_num);
1316 port_s->enable = false;
1319 bitmap_set1(port_num_bitmap, port_idx);
1320 port_s->port_num = port_idx;
1322 if (*port_num_counter >= STP_MAX_PORTS) {
1323 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1324 port_s->enable = false;
1328 port_s->port_num = (*port_num_counter)++;
1331 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1333 port_s->path_cost = strtoul(config_str, NULL, 10);
1335 enum netdev_features current;
1338 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1339 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1340 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1343 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1345 port_s->priority = strtoul(config_str, NULL, 0);
1347 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1352 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1353 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1355 const char *config_str;
1356 struct iface *iface;
1358 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1359 port_s->enable = false;
1362 port_s->enable = true;
1365 /* RSTP over bonds is not supported. */
1366 if (!list_is_singleton(&port->ifaces)) {
1367 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1369 port_s->enable = false;
1373 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1375 /* Internal ports shouldn't participate in spanning tree, so
1377 if (!strcmp(iface->type, "internal")) {
1378 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1379 port_s->enable = false;
1383 /* RSTP on mirror output ports is not supported. */
1384 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1385 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1386 port_s->enable = false;
1390 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1392 unsigned long int port_num = strtoul(config_str, NULL, 0);
1393 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1394 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1395 port_s->enable = false;
1398 port_s->port_num = port_num;
1400 if (*port_num_counter >= RSTP_MAX_PORTS) {
1401 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1402 port_s->enable = false;
1405 /* If rstp-port-num is not specified, use 0.
1406 * rstp_port_set_port_number() will look for the first free one. */
1407 port_s->port_num = 0;
1410 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1412 port_s->path_cost = strtoul(config_str, NULL, 10);
1414 enum netdev_features current;
1417 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1418 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1419 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1422 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1424 port_s->priority = strtoul(config_str, NULL, 0);
1426 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1429 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1431 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1433 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1436 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1437 "rstp-admin-port-state", true);
1439 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1440 "rstp-port-admin-edge", false);
1441 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1442 "rstp-port-auto-edge", true);
1443 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1444 "rstp-port-mcheck", false);
1447 /* Set spanning tree configuration on 'br'. */
1449 bridge_configure_stp(struct bridge *br, bool enable_stp)
1452 ofproto_set_stp(br->ofproto, NULL);
1454 struct ofproto_stp_settings br_s;
1455 const char *config_str;
1457 int port_num_counter;
1458 unsigned long *port_num_bitmap;
1460 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1462 uint8_t ea[ETH_ADDR_LEN];
1464 if (eth_addr_from_string(config_str, ea)) {
1465 br_s.system_id = eth_addr_to_uint64(ea);
1467 br_s.system_id = eth_addr_to_uint64(br->ea);
1468 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1469 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1472 br_s.system_id = eth_addr_to_uint64(br->ea);
1475 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1477 br_s.priority = strtoul(config_str, NULL, 0);
1479 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1482 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1484 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1486 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1489 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1491 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1493 br_s.max_age = STP_DEFAULT_MAX_AGE;
1496 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1498 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1500 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1503 /* Configure STP on the bridge. */
1504 if (ofproto_set_stp(br->ofproto, &br_s)) {
1505 VLOG_ERR("bridge %s: could not enable STP", br->name);
1509 /* Users must either set the port number with the "stp-port-num"
1510 * configuration on all ports or none. If manual configuration
1511 * is not done, then we allocate them sequentially. */
1512 port_num_counter = 0;
1513 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1514 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1515 struct ofproto_port_stp_settings port_s;
1516 struct iface *iface;
1518 port_configure_stp(br->ofproto, port, &port_s,
1519 &port_num_counter, port_num_bitmap);
1521 /* As bonds are not supported, just apply configuration to
1522 * all interfaces. */
1523 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1524 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1526 VLOG_ERR("port %s: could not enable STP", port->name);
1532 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1533 && port_num_counter) {
1534 VLOG_ERR("bridge %s: must manually configure all STP port "
1535 "IDs or none, disabling", br->name);
1536 ofproto_set_stp(br->ofproto, NULL);
1538 bitmap_free(port_num_bitmap);
1543 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1546 ofproto_set_rstp(br->ofproto, NULL);
1548 struct ofproto_rstp_settings br_s;
1549 const char *config_str;
1551 int port_num_counter;
1553 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1555 uint8_t ea[ETH_ADDR_LEN];
1557 if (eth_addr_from_string(config_str, ea)) {
1558 br_s.address = eth_addr_to_uint64(ea);
1561 br_s.address = eth_addr_to_uint64(br->ea);
1562 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1563 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1567 br_s.address = eth_addr_to_uint64(br->ea);
1570 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1572 br_s.priority = strtoul(config_str, NULL, 0);
1574 br_s.priority = RSTP_DEFAULT_PRIORITY;
1577 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1579 br_s.ageing_time = strtoul(config_str, NULL, 0);
1581 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1584 config_str = smap_get(&br->cfg->other_config,
1585 "rstp-force-protocol-version");
1587 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1589 br_s.force_protocol_version = FPV_DEFAULT;
1592 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1594 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1596 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1599 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1601 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1603 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1606 config_str = smap_get(&br->cfg->other_config,
1607 "rstp-transmit-hold-count");
1609 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1611 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1614 /* Configure RSTP on the bridge. */
1615 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1616 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1620 port_num_counter = 0;
1621 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1622 struct ofproto_port_rstp_settings port_s;
1623 struct iface *iface;
1625 port_configure_rstp(br->ofproto, port, &port_s,
1628 /* As bonds are not supported, just apply configuration to
1629 * all interfaces. */
1630 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1631 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1633 VLOG_ERR("port %s: could not enable RSTP", port->name);
1642 bridge_configure_spanning_tree(struct bridge *br)
1644 bool enable_rstp = br->cfg->rstp_enable;
1645 bool enable_stp = br->cfg->stp_enable;
1647 if (enable_rstp && enable_stp) {
1648 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1649 "configured; enabling RSTP", br->name);
1653 bridge_configure_stp(br, enable_stp);
1654 bridge_configure_rstp(br, enable_rstp);
1658 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1660 const struct port *port = port_lookup(br, name);
1661 return port && port_is_bond_fake_iface(port);
1665 port_is_bond_fake_iface(const struct port *port)
1667 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1671 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1673 struct bridge *br, *next;
1674 struct shash new_br;
1677 /* Collect new bridges' names and types. */
1678 shash_init(&new_br);
1679 for (i = 0; i < cfg->n_bridges; i++) {
1680 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1681 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1683 if (strchr(br_cfg->name, '/')) {
1684 /* Prevent remote ovsdb-server users from accessing arbitrary
1685 * directories, e.g. consider a bridge named "../../../etc/". */
1686 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1688 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1689 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1693 /* Get rid of deleted bridges or those whose types have changed.
1694 * Update 'cfg' of bridges that still exist. */
1695 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1696 br->cfg = shash_find_data(&new_br, br->name);
1697 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1698 br->cfg->datapath_type))) {
1703 /* Add new bridges. */
1704 for (i = 0; i < cfg->n_bridges; i++) {
1705 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1706 struct bridge *br = bridge_lookup(br_cfg->name);
1708 bridge_create(br_cfg);
1712 shash_destroy(&new_br);
1715 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1716 * Returns 0 if successful, otherwise a positive errno value. */
1718 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1719 struct netdev *netdev, char **errp)
1721 return netdev_set_config(netdev, &iface_cfg->options, errp);
1724 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1725 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1727 * If successful, returns 0 and stores the network device in '*netdevp'. On
1728 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1730 iface_do_create(const struct bridge *br,
1731 const struct ovsrec_interface *iface_cfg,
1732 const struct ovsrec_port *port_cfg,
1733 ofp_port_t *ofp_portp, struct netdev **netdevp,
1736 struct netdev *netdev = NULL;
1739 if (netdev_is_reserved_name(iface_cfg->name)) {
1740 VLOG_WARN("could not create interface %s, name is reserved",
1746 error = netdev_open(iface_cfg->name,
1747 iface_get_type(iface_cfg, br->cfg), &netdev);
1749 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1750 iface_cfg->name, ovs_strerror(error));
1754 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1759 *ofp_portp = iface_pick_ofport(iface_cfg);
1760 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1765 VLOG_INFO("bridge %s: added interface %s on port %d",
1766 br->name, iface_cfg->name, *ofp_portp);
1768 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1769 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1777 netdev_close(netdev);
1781 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1782 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1783 * automatically allocated for the iface. Takes ownership of and
1784 * deallocates 'if_cfg'.
1786 * Return true if an iface is successfully created, false otherwise. */
1788 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1789 const struct ovsrec_port *port_cfg)
1791 struct netdev *netdev;
1792 struct iface *iface;
1793 ofp_port_t ofp_port;
1798 /* Do the bits that can fail up front. */
1799 ovs_assert(!iface_lookup(br, iface_cfg->name));
1800 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1802 iface_clear_db_record(iface_cfg, errp);
1807 /* Get or create the port structure. */
1808 port = port_lookup(br, port_cfg->name);
1810 port = port_create(br, port_cfg);
1813 /* Create the iface structure. */
1814 iface = xzalloc(sizeof *iface);
1815 list_push_back(&port->ifaces, &iface->port_elem);
1816 hmap_insert(&br->iface_by_name, &iface->name_node,
1817 hash_string(iface_cfg->name, 0));
1819 iface->name = xstrdup(iface_cfg->name);
1820 iface->ofp_port = ofp_port;
1821 iface->netdev = netdev;
1822 iface->type = iface_get_type(iface_cfg, br->cfg);
1823 iface->cfg = iface_cfg;
1824 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1825 hash_ofp_port(ofp_port));
1827 /* Populate initial status in database. */
1828 iface_refresh_stats(iface);
1829 iface_refresh_netdev_status(iface);
1831 /* Add bond fake iface if necessary. */
1832 if (port_is_bond_fake_iface(port)) {
1833 struct ofproto_port ofproto_port;
1835 if (ofproto_port_query_by_name(br->ofproto, port->name,
1837 struct netdev *netdev;
1840 error = netdev_open(port->name, "internal", &netdev);
1842 ofp_port_t fake_ofp_port = OFPP_NONE;
1843 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1844 netdev_close(netdev);
1846 VLOG_WARN("could not open network device %s (%s)",
1847 port->name, ovs_strerror(error));
1850 /* Already exists, nothing to do. */
1851 ofproto_port_destroy(&ofproto_port);
1858 /* Set forward BPDU option. */
1860 bridge_configure_forward_bpdu(struct bridge *br)
1862 ofproto_set_forward_bpdu(br->ofproto,
1863 smap_get_bool(&br->cfg->other_config,
1868 /* Set MAC learning table configuration for 'br'. */
1870 bridge_configure_mac_table(struct bridge *br)
1872 const char *idle_time_str;
1875 const char *mac_table_size_str;
1878 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1879 idle_time = (idle_time_str && atoi(idle_time_str)
1880 ? atoi(idle_time_str)
1881 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1883 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1884 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1885 ? atoi(mac_table_size_str)
1888 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1891 /* Set multicast snooping table configuration for 'br'. */
1893 bridge_configure_mcast_snooping(struct bridge *br)
1895 if (!br->cfg->mcast_snooping_enable) {
1896 ofproto_set_mcast_snooping(br->ofproto, NULL);
1899 struct ofproto_mcast_snooping_settings br_s;
1900 const char *idle_time_str;
1901 const char *max_entries_str;
1903 idle_time_str = smap_get(&br->cfg->other_config,
1904 "mcast-snooping-aging-time");
1905 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1906 ? atoi(idle_time_str)
1907 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1909 max_entries_str = smap_get(&br->cfg->other_config,
1910 "mcast-snooping-table-size");
1911 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1912 ? atoi(max_entries_str)
1913 : MCAST_DEFAULT_MAX_ENTRIES);
1915 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1916 "mcast-snooping-disable-flood-unregistered",
1919 /* Configure multicast snooping on the bridge */
1920 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1921 VLOG_ERR("bridge %s: could not enable multicast snooping",
1926 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1927 struct ofproto_mcast_snooping_port_settings port_s;
1928 port_s.flood = smap_get_bool(&port->cfg->other_config,
1929 "mcast-snooping-flood", false);
1930 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1931 "mcast-snooping-flood-reports", false);
1932 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1933 VLOG_ERR("port %s: could not configure mcast snooping",
1941 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1942 const struct port *fake_br, struct iface **hw_addr_iface)
1944 struct hmapx mirror_output_ports;
1946 bool found_addr = false;
1950 /* Mirror output ports don't participate in picking the local hardware
1951 * address. ofproto can't help us find out whether a given port is a
1952 * mirror output because we haven't configured mirrors yet, so we need to
1953 * accumulate them ourselves. */
1954 hmapx_init(&mirror_output_ports);
1955 for (i = 0; i < br->cfg->n_mirrors; i++) {
1956 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1957 if (m->output_port) {
1958 hmapx_add(&mirror_output_ports, m->output_port);
1962 /* Otherwise choose the minimum non-local MAC address among all of the
1964 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1965 uint8_t iface_ea[ETH_ADDR_LEN];
1966 struct iface *candidate;
1967 struct iface *iface;
1969 /* Mirror output ports don't participate. */
1970 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1974 /* Choose the MAC address to represent the port. */
1976 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1977 /* Find the interface with this Ethernet address (if any) so that
1978 * we can provide the correct devname to the caller. */
1979 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1980 uint8_t candidate_ea[ETH_ADDR_LEN];
1981 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1982 && eth_addr_equals(iface_ea, candidate_ea)) {
1987 /* Choose the interface whose MAC address will represent the port.
1988 * The Linux kernel bonding code always chooses the MAC address of
1989 * the first slave added to a bond, and the Fedora networking
1990 * scripts always add slaves to a bond in alphabetical order, so
1991 * for compatibility we choose the interface with the name that is
1992 * first in alphabetical order. */
1993 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1994 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1999 /* The local port doesn't count (since we're trying to choose its
2000 * MAC address anyway). */
2001 if (iface->ofp_port == OFPP_LOCAL) {
2005 /* For fake bridges we only choose from ports with the same tag */
2006 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2007 if (!port->cfg->tag) {
2010 if (*port->cfg->tag != *fake_br->cfg->tag) {
2016 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2022 /* Compare against our current choice. */
2023 if (!eth_addr_is_multicast(iface_ea) &&
2024 !eth_addr_is_local(iface_ea) &&
2025 !eth_addr_is_reserved(iface_ea) &&
2026 !eth_addr_is_zero(iface_ea) &&
2027 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2029 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2030 *hw_addr_iface = iface;
2036 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2037 *hw_addr_iface = NULL;
2040 hmapx_destroy(&mirror_output_ports);
2044 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2045 struct iface **hw_addr_iface)
2048 *hw_addr_iface = NULL;
2050 /* Did the user request a particular MAC? */
2051 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2052 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2053 if (eth_addr_is_multicast(ea)) {
2054 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2055 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2056 } else if (eth_addr_is_zero(ea)) {
2057 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2063 /* Find a local hw address */
2064 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2067 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2068 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2069 * an interface on 'br', then that interface must be passed in as
2070 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2071 * 'hw_addr_iface' must be passed in as a null pointer. */
2073 bridge_pick_datapath_id(struct bridge *br,
2074 const uint8_t bridge_ea[ETH_ADDR_LEN],
2075 struct iface *hw_addr_iface)
2078 * The procedure for choosing a bridge MAC address will, in the most
2079 * ordinary case, also choose a unique MAC that we can use as a datapath
2080 * ID. In some special cases, though, multiple bridges will end up with
2081 * the same MAC address. This is OK for the bridges, but it will confuse
2082 * the OpenFlow controller, because each datapath needs a unique datapath
2085 * Datapath IDs must be unique. It is also very desirable that they be
2086 * stable from one run to the next, so that policy set on a datapath
2089 const char *datapath_id;
2092 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2093 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2097 if (!hw_addr_iface) {
2099 * A purely internal bridge, that is, one that has no non-virtual
2100 * network devices on it at all, is difficult because it has no
2101 * natural unique identifier at all.
2103 * When the host is a XenServer, we handle this case by hashing the
2104 * host's UUID with the name of the bridge. Names of bridges are
2105 * persistent across XenServer reboots, although they can be reused if
2106 * an internal network is destroyed and then a new one is later
2107 * created, so this is fairly effective.
2109 * When the host is not a XenServer, we punt by using a random MAC
2110 * address on each run.
2112 const char *host_uuid = xenserver_get_host_uuid();
2114 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2115 dpid = dpid_from_hash(combined, strlen(combined));
2121 return eth_addr_to_uint64(bridge_ea);
2125 dpid_from_hash(const void *data, size_t n)
2127 uint8_t hash[SHA1_DIGEST_SIZE];
2129 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2130 sha1_bytes(data, n, hash);
2131 eth_addr_mark_random(hash);
2132 return eth_addr_to_uint64(hash);
2136 iface_refresh_netdev_status(struct iface *iface)
2140 enum netdev_features current;
2141 enum netdev_flags flags;
2142 const char *link_state;
2143 uint8_t mac[ETH_ADDR_LEN];
2144 int64_t bps, mtu_64, ifindex64, link_resets;
2147 if (iface_is_synthetic(iface)) {
2151 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2152 && !status_txn_try_again) {
2156 iface->change_seq = netdev_get_change_seq(iface->netdev);
2160 if (!netdev_get_status(iface->netdev, &smap)) {
2161 ovsrec_interface_set_status(iface->cfg, &smap);
2163 ovsrec_interface_set_status(iface->cfg, NULL);
2166 smap_destroy(&smap);
2168 error = netdev_get_flags(iface->netdev, &flags);
2170 const char *state = flags & NETDEV_UP ? "up" : "down";
2172 ovsrec_interface_set_admin_state(iface->cfg, state);
2174 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2177 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2178 ovsrec_interface_set_link_state(iface->cfg, link_state);
2180 link_resets = netdev_get_carrier_resets(iface->netdev);
2181 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2183 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2184 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2186 ovsrec_interface_set_duplex(iface->cfg,
2187 netdev_features_is_full_duplex(current)
2189 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2191 ovsrec_interface_set_duplex(iface->cfg, NULL);
2192 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2195 error = netdev_get_mtu(iface->netdev, &mtu);
2198 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2200 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2203 error = netdev_get_etheraddr(iface->netdev, mac);
2205 char mac_string[32];
2207 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2208 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2210 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2213 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2214 * if there is no valid ifindex number. */
2215 ifindex64 = netdev_get_ifindex(iface->netdev);
2216 if (ifindex64 < 0) {
2219 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2223 iface_refresh_ofproto_status(struct iface *iface)
2227 if (iface_is_synthetic(iface)) {
2231 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2235 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2237 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2240 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2242 || status_txn_try_again) {
2243 iface_refresh_cfm_stats(iface);
2246 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2248 || status_txn_try_again) {
2252 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2253 iface->ofp_port, &smap);
2254 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2255 smap_destroy(&smap);
2259 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2262 iface_refresh_cfm_stats(struct iface *iface)
2264 const struct ovsrec_interface *cfg = iface->cfg;
2265 struct cfm_status status;
2268 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2269 iface->ofp_port, &status);
2271 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2272 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2273 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2274 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2275 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2276 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2278 const char *reasons[CFM_FAULT_N_REASONS];
2279 int64_t cfm_health = status.health;
2280 int64_t cfm_flap_count = status.flap_count;
2281 bool faulted = status.faults != 0;
2284 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2287 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2288 int reason = 1 << i;
2289 if (status.faults & reason) {
2290 reasons[j++] = cfm_fault_reason_to_str(reason);
2293 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2295 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2297 if (status.remote_opstate >= 0) {
2298 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2299 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2301 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2304 ovsrec_interface_set_cfm_remote_mpids(cfg,
2305 (const int64_t *)status.rmps,
2307 if (cfm_health >= 0) {
2308 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2310 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2318 iface_refresh_stats(struct iface *iface)
2320 #define IFACE_STATS \
2321 IFACE_STAT(rx_packets, "rx_packets") \
2322 IFACE_STAT(tx_packets, "tx_packets") \
2323 IFACE_STAT(rx_bytes, "rx_bytes") \
2324 IFACE_STAT(tx_bytes, "tx_bytes") \
2325 IFACE_STAT(rx_dropped, "rx_dropped") \
2326 IFACE_STAT(tx_dropped, "tx_dropped") \
2327 IFACE_STAT(rx_errors, "rx_errors") \
2328 IFACE_STAT(tx_errors, "tx_errors") \
2329 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2330 IFACE_STAT(rx_over_errors, "rx_over_err") \
2331 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2332 IFACE_STAT(collisions, "collisions")
2334 #define IFACE_STAT(MEMBER, NAME) + 1
2335 enum { N_IFACE_STATS = IFACE_STATS };
2337 int64_t values[N_IFACE_STATS];
2338 const char *keys[N_IFACE_STATS];
2341 struct netdev_stats stats;
2343 if (iface_is_synthetic(iface)) {
2347 /* Intentionally ignore return value, since errors will set 'stats' to
2348 * all-1s, and we will deal with that correctly below. */
2349 netdev_get_stats(iface->netdev, &stats);
2351 /* Copy statistics into keys[] and values[]. */
2353 #define IFACE_STAT(MEMBER, NAME) \
2354 if (stats.MEMBER != UINT64_MAX) { \
2356 values[n] = stats.MEMBER; \
2361 ovs_assert(n <= N_IFACE_STATS);
2363 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2368 br_refresh_datapath_info(struct bridge *br)
2370 const char *version;
2372 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2373 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2376 ovsrec_bridge_set_datapath_version(br->cfg,
2377 version ? version : "<unknown>");
2381 br_refresh_stp_status(struct bridge *br)
2383 struct smap smap = SMAP_INITIALIZER(&smap);
2384 struct ofproto *ofproto = br->ofproto;
2385 struct ofproto_stp_status status;
2387 if (ofproto_get_stp_status(ofproto, &status)) {
2391 if (!status.enabled) {
2392 ovsrec_bridge_set_status(br->cfg, NULL);
2396 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2397 STP_ID_ARGS(status.bridge_id));
2398 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2399 STP_ID_ARGS(status.designated_root));
2400 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2402 ovsrec_bridge_set_status(br->cfg, &smap);
2403 smap_destroy(&smap);
2407 port_refresh_stp_status(struct port *port)
2409 struct ofproto *ofproto = port->bridge->ofproto;
2410 struct iface *iface;
2411 struct ofproto_port_stp_status status;
2414 if (port_is_synthetic(port)) {
2418 /* STP doesn't currently support bonds. */
2419 if (!list_is_singleton(&port->ifaces)) {
2420 ovsrec_port_set_status(port->cfg, NULL);
2424 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2425 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2429 if (!status.enabled) {
2430 ovsrec_port_set_status(port->cfg, NULL);
2434 /* Set Status column. */
2436 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2437 smap_add(&smap, "stp_state", stp_state_name(status.state));
2438 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2439 smap_add(&smap, "stp_role", stp_role_name(status.role));
2440 ovsrec_port_set_status(port->cfg, &smap);
2441 smap_destroy(&smap);
2445 port_refresh_stp_stats(struct port *port)
2447 struct ofproto *ofproto = port->bridge->ofproto;
2448 struct iface *iface;
2449 struct ofproto_port_stp_stats stats;
2450 const char *keys[3];
2451 int64_t int_values[3];
2453 if (port_is_synthetic(port)) {
2457 /* STP doesn't currently support bonds. */
2458 if (!list_is_singleton(&port->ifaces)) {
2462 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2463 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2467 if (!stats.enabled) {
2468 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2472 /* Set Statistics column. */
2473 keys[0] = "stp_tx_count";
2474 int_values[0] = stats.tx_count;
2475 keys[1] = "stp_rx_count";
2476 int_values[1] = stats.rx_count;
2477 keys[2] = "stp_error_count";
2478 int_values[2] = stats.error_count;
2480 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2481 ARRAY_SIZE(int_values));
2485 br_refresh_rstp_status(struct bridge *br)
2487 struct smap smap = SMAP_INITIALIZER(&smap);
2488 struct ofproto *ofproto = br->ofproto;
2489 struct ofproto_rstp_status status;
2491 if (ofproto_get_rstp_status(ofproto, &status)) {
2494 if (!status.enabled) {
2495 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2498 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2499 RSTP_ID_ARGS(status.bridge_id));
2500 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2501 status.root_path_cost);
2502 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2503 RSTP_ID_ARGS(status.root_id));
2504 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2505 RSTP_ID_ARGS(status.designated_id));
2506 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2507 status.designated_port_id);
2508 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2509 status.bridge_port_id);
2510 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2511 smap_destroy(&smap);
2515 port_refresh_rstp_status(struct port *port)
2517 struct ofproto *ofproto = port->bridge->ofproto;
2518 struct iface *iface;
2519 struct ofproto_port_rstp_status status;
2520 const char *keys[4];
2521 int64_t int_values[4];
2524 if (port_is_synthetic(port)) {
2528 /* RSTP doesn't currently support bonds. */
2529 if (!list_is_singleton(&port->ifaces)) {
2530 ovsrec_port_set_rstp_status(port->cfg, NULL);
2534 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2535 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2539 if (!status.enabled) {
2540 ovsrec_port_set_rstp_status(port->cfg, NULL);
2541 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2544 /* Set Status column. */
2547 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2549 smap_add_format(&smap, "rstp_port_role", "%s",
2550 rstp_port_role_name(status.role));
2551 smap_add_format(&smap, "rstp_port_state", "%s",
2552 rstp_state_name(status.state));
2553 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2554 RSTP_ID_ARGS(status.designated_bridge_id));
2555 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2556 status.designated_port_id);
2557 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2558 status.designated_path_cost);
2560 ovsrec_port_set_rstp_status(port->cfg, &smap);
2561 smap_destroy(&smap);
2563 /* Set Statistics column. */
2564 keys[0] = "rstp_tx_count";
2565 int_values[0] = status.tx_count;
2566 keys[1] = "rstp_rx_count";
2567 int_values[1] = status.rx_count;
2568 keys[2] = "rstp_uptime";
2569 int_values[2] = status.uptime;
2570 keys[3] = "rstp_error_count";
2571 int_values[3] = status.error_count;
2572 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2573 ARRAY_SIZE(int_values));
2577 port_refresh_bond_status(struct port *port, bool force_update)
2579 uint8_t mac[ETH_ADDR_LEN];
2581 /* Return if port is not a bond */
2582 if (list_is_singleton(&port->ifaces)) {
2586 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2590 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2591 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2597 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2599 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2603 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2605 bool enable = enable_system_stats(cfg);
2607 system_stats_enable(enable);
2609 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2614 run_system_stats(void)
2616 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2619 stats = system_stats_run();
2621 struct ovsdb_idl_txn *txn;
2622 struct ovsdb_datum datum;
2624 txn = ovsdb_idl_txn_create(idl);
2625 ovsdb_datum_from_smap(&datum, stats);
2626 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2628 ovsdb_idl_txn_commit(txn);
2629 ovsdb_idl_txn_destroy(txn);
2636 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2639 case OFPCR12_ROLE_EQUAL:
2641 case OFPCR12_ROLE_MASTER:
2643 case OFPCR12_ROLE_SLAVE:
2645 case OFPCR12_ROLE_NOCHANGE:
2647 return "*** INVALID ROLE ***";
2652 refresh_controller_status(void)
2656 const struct ovsrec_controller *cfg;
2660 /* Accumulate status for controllers on all bridges. */
2661 HMAP_FOR_EACH (br, node, &all_bridges) {
2662 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2665 /* Update each controller in the database with current status. */
2666 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2667 struct ofproto_controller_info *cinfo =
2668 shash_find_data(&info, cfg->target);
2671 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2672 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2674 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2676 ovsrec_controller_set_is_connected(cfg, false);
2677 ovsrec_controller_set_role(cfg, NULL);
2678 ovsrec_controller_set_status(cfg, NULL);
2682 ofproto_free_ofproto_controller_info(&info);
2685 /* Update interface and mirror statistics if necessary. */
2687 run_stats_update(void)
2689 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2696 /* Statistics update interval should always be greater than or equal to
2698 stats_interval = MAX(smap_get_int(&cfg->other_config,
2699 "stats-update-interval",
2701 if (stats_timer_interval != stats_interval) {
2702 stats_timer_interval = stats_interval;
2703 stats_timer = LLONG_MIN;
2706 if (time_msec() >= stats_timer) {
2707 enum ovsdb_idl_txn_status status;
2709 /* Rate limit the update. Do not start a new update if the
2710 * previous one is not done. */
2714 stats_txn = ovsdb_idl_txn_create(idl);
2715 HMAP_FOR_EACH (br, node, &all_bridges) {
2719 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2720 struct iface *iface;
2722 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2723 iface_refresh_stats(iface);
2725 port_refresh_stp_stats(port);
2727 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2728 mirror_refresh_stats(m);
2731 refresh_controller_status();
2734 status = ovsdb_idl_txn_commit(stats_txn);
2735 if (status != TXN_INCOMPLETE) {
2736 stats_timer = time_msec() + stats_timer_interval;
2737 ovsdb_idl_txn_destroy(stats_txn);
2744 stats_update_wait(void)
2746 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2747 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2749 ovsdb_idl_txn_wait(stats_txn);
2751 poll_timer_wait_until(stats_timer);
2755 /* Update bridge/port/interface status if necessary. */
2757 run_status_update(void)
2762 /* Rate limit the update. Do not start a new update if the
2763 * previous one is not done. */
2764 seq = seq_read(connectivity_seq_get());
2765 if (seq != connectivity_seqno || status_txn_try_again) {
2768 connectivity_seqno = seq;
2769 status_txn = ovsdb_idl_txn_create(idl);
2770 HMAP_FOR_EACH (br, node, &all_bridges) {
2773 br_refresh_stp_status(br);
2774 br_refresh_rstp_status(br);
2775 br_refresh_datapath_info(br);
2776 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2777 struct iface *iface;
2779 port_refresh_stp_status(port);
2780 port_refresh_rstp_status(port);
2781 port_refresh_bond_status(port, status_txn_try_again);
2782 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2783 iface_refresh_netdev_status(iface);
2784 iface_refresh_ofproto_status(iface);
2791 /* Commit the transaction and get the status. If the transaction finishes,
2792 * then destroy the transaction. Otherwise, keep it so that we can check
2793 * progress the next time that this function is called. */
2795 enum ovsdb_idl_txn_status status;
2797 status = ovsdb_idl_txn_commit(status_txn);
2798 if (status != TXN_INCOMPLETE) {
2799 ovsdb_idl_txn_destroy(status_txn);
2802 /* Sets the 'status_txn_try_again' if the transaction fails. */
2803 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2804 status_txn_try_again = false;
2806 status_txn_try_again = true;
2811 /* Refresh AA port status if necessary. */
2812 if (time_msec() >= aa_refresh_timer) {
2815 HMAP_FOR_EACH (br, node, &all_bridges) {
2816 if (bridge_aa_need_refresh(br)) {
2817 struct ovsdb_idl_txn *txn;
2819 txn = ovsdb_idl_txn_create(idl);
2820 bridge_aa_refresh_queued(br);
2821 ovsdb_idl_txn_commit(txn);
2822 ovsdb_idl_txn_destroy(txn);
2826 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2831 status_update_wait(void)
2833 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2834 * transaction to complete. If the status update to database needs to be
2835 * run again (transaction fails), registers a timeout in
2836 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2837 * sequence number. */
2839 ovsdb_idl_txn_wait(status_txn);
2840 } else if (status_txn_try_again) {
2841 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2843 seq_wait(connectivity_seq_get(), connectivity_seqno);
2854 /* Let each datapath type do the work that it needs to do. */
2856 ofproto_enumerate_types(&types);
2857 SSET_FOR_EACH (type, &types) {
2858 ofproto_type_run(type);
2860 sset_destroy(&types);
2862 /* Let each bridge do the work that it needs to do. */
2863 HMAP_FOR_EACH (br, node, &all_bridges) {
2864 ofproto_run(br->ofproto);
2871 static struct ovsrec_open_vswitch null_cfg;
2872 const struct ovsrec_open_vswitch *cfg;
2874 bool vlan_splinters_changed;
2876 ovsrec_open_vswitch_init(&null_cfg);
2880 if (ovsdb_idl_is_lock_contended(idl)) {
2881 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2882 struct bridge *br, *next_br;
2884 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2885 "disabling this process (pid %ld) until it goes away",
2886 (long int) getpid());
2888 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2891 /* Since we will not be running system_stats_run() in this process
2892 * with the current situation of multiple ovs-vswitchd daemons,
2893 * disable system stats collection. */
2894 system_stats_enable(false);
2896 } else if (!ovsdb_idl_has_lock(idl)
2897 || !ovsdb_idl_has_ever_connected(idl)) {
2898 /* Returns if not holding the lock or not done retrieving db
2902 cfg = ovsrec_open_vswitch_first(idl);
2904 /* Initialize the ofproto library. This only needs to run once, but
2905 * it must be done after the configuration is set. If the
2906 * initialization has already occurred, bridge_init_ofproto()
2907 * returns immediately. */
2908 bridge_init_ofproto(cfg);
2910 /* Once the value of flow-restore-wait is false, we no longer should
2911 * check its value from the database. */
2912 if (cfg && ofproto_get_flow_restore_wait()) {
2913 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2914 "flow-restore-wait", false));
2919 /* Re-configure SSL. We do this on every trip through the main loop,
2920 * instead of just when the database changes, because the contents of the
2921 * key and certificate files can change without the database changing.
2923 * We do this before bridge_reconfigure() because that function might
2924 * initiate SSL connections and thus requires SSL to be configured. */
2925 if (cfg && cfg->ssl) {
2926 const struct ovsrec_ssl *ssl = cfg->ssl;
2928 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2929 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2932 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2933 * usage has changed. */
2934 vlan_splinters_changed = false;
2935 if (vlan_splinters_enabled_anywhere) {
2938 HMAP_FOR_EACH (br, node, &all_bridges) {
2939 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2940 vlan_splinters_changed = true;
2946 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2947 struct ovsdb_idl_txn *txn;
2949 idl_seqno = ovsdb_idl_get_seqno(idl);
2950 txn = ovsdb_idl_txn_create(idl);
2951 bridge_reconfigure(cfg ? cfg : &null_cfg);
2954 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2955 discover_types(cfg);
2958 /* If we are completing our initial configuration for this run
2959 * of ovs-vswitchd, then keep the transaction around to monitor
2960 * it for completion. */
2961 if (initial_config_done) {
2962 /* Always sets the 'status_txn_try_again' to check again,
2963 * in case that this transaction fails. */
2964 status_txn_try_again = true;
2965 ovsdb_idl_txn_commit(txn);
2966 ovsdb_idl_txn_destroy(txn);
2968 initial_config_done = true;
2969 daemonize_txn = txn;
2973 if (daemonize_txn) {
2974 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2975 if (status != TXN_INCOMPLETE) {
2976 ovsdb_idl_txn_destroy(daemonize_txn);
2977 daemonize_txn = NULL;
2979 /* ovs-vswitchd has completed initialization, so allow the
2980 * process that forked us to exit successfully. */
2981 daemonize_complete();
2983 vlog_enable_async();
2985 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2990 run_status_update();
3000 ovsdb_idl_wait(idl);
3001 if (daemonize_txn) {
3002 ovsdb_idl_txn_wait(daemonize_txn);
3006 ofproto_enumerate_types(&types);
3007 SSET_FOR_EACH (type, &types) {
3008 ofproto_type_wait(type);
3010 sset_destroy(&types);
3012 if (!hmap_is_empty(&all_bridges)) {
3015 HMAP_FOR_EACH (br, node, &all_bridges) {
3016 ofproto_wait(br->ofproto);
3018 stats_update_wait();
3019 status_update_wait();
3022 system_stats_wait();
3025 /* Adds some memory usage statistics for bridges into 'usage', for use with
3026 * memory_report(). */
3028 bridge_get_memory_usage(struct simap *usage)
3035 ofproto_enumerate_types(&types);
3036 SSET_FOR_EACH (type, &types) {
3037 ofproto_type_get_memory_usage(type, usage);
3039 sset_destroy(&types);
3041 HMAP_FOR_EACH (br, node, &all_bridges) {
3042 ofproto_get_memory_usage(br->ofproto, usage);
3046 /* QoS unixctl user interface functions. */
3048 struct qos_unixctl_show_cbdata {
3050 struct iface *iface;
3054 qos_unixctl_show_queue(unsigned int queue_id,
3055 const struct smap *details,
3056 struct iface *iface,
3059 struct netdev_queue_stats stats;
3060 struct smap_node *node;
3063 ds_put_cstr(ds, "\n");
3065 ds_put_format(ds, "Queue %u:\n", queue_id);
3067 ds_put_cstr(ds, "Default:\n");
3070 SMAP_FOR_EACH (node, details) {
3071 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3074 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3076 if (stats.tx_packets != UINT64_MAX) {
3077 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3080 if (stats.tx_bytes != UINT64_MAX) {
3081 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3084 if (stats.tx_errors != UINT64_MAX) {
3085 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3088 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3089 queue_id, ovs_strerror(error));
3094 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3095 const char *argv[], void *aux OVS_UNUSED)
3097 struct ds ds = DS_EMPTY_INITIALIZER;
3098 struct smap smap = SMAP_INITIALIZER(&smap);
3099 struct iface *iface;
3101 struct smap_node *node;
3103 iface = iface_find(argv[1]);
3105 unixctl_command_reply_error(conn, "no such interface");
3109 netdev_get_qos(iface->netdev, &type, &smap);
3111 if (*type != '\0') {
3112 struct netdev_queue_dump dump;
3113 struct smap details;
3114 unsigned int queue_id;
3116 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3118 SMAP_FOR_EACH (node, &smap) {
3119 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3122 smap_init(&details);
3123 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3124 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3126 smap_destroy(&details);
3128 unixctl_command_reply(conn, ds_cstr(&ds));
3130 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3131 unixctl_command_reply_error(conn, ds_cstr(&ds));
3134 smap_destroy(&smap);
3138 /* Bridge reconfiguration functions. */
3140 bridge_create(const struct ovsrec_bridge *br_cfg)
3144 ovs_assert(!bridge_lookup(br_cfg->name));
3145 br = xzalloc(sizeof *br);
3147 br->name = xstrdup(br_cfg->name);
3148 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3151 /* Derive the default Ethernet address from the bridge's UUID. This should
3152 * be unique and it will be stable between ovs-vswitchd runs. */
3153 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3154 eth_addr_mark_random(br->default_ea);
3156 hmap_init(&br->ports);
3157 hmap_init(&br->ifaces);
3158 hmap_init(&br->iface_by_name);
3159 hmap_init(&br->mirrors);
3161 hmap_init(&br->mappings);
3162 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3166 bridge_destroy(struct bridge *br)
3169 struct mirror *mirror, *next_mirror;
3170 struct port *port, *next_port;
3172 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3175 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3176 mirror_destroy(mirror);
3179 hmap_remove(&all_bridges, &br->node);
3180 ofproto_destroy(br->ofproto);
3181 hmap_destroy(&br->ifaces);
3182 hmap_destroy(&br->ports);
3183 hmap_destroy(&br->iface_by_name);
3184 hmap_destroy(&br->mirrors);
3185 hmap_destroy(&br->mappings);
3192 static struct bridge *
3193 bridge_lookup(const char *name)
3197 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3198 if (!strcmp(br->name, name)) {
3205 /* Handle requests for a listing of all flows known by the OpenFlow
3206 * stack, including those normally hidden. */
3208 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3209 const char *argv[], void *aux OVS_UNUSED)
3214 br = bridge_lookup(argv[1]);
3216 unixctl_command_reply_error(conn, "Unknown bridge");
3221 ofproto_get_all_flows(br->ofproto, &results);
3223 unixctl_command_reply(conn, ds_cstr(&results));
3224 ds_destroy(&results);
3227 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3228 * connections and reconnect. If BRIDGE is not specified, then all bridges
3229 * drop their controller connections and reconnect. */
3231 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3232 const char *argv[], void *aux OVS_UNUSED)
3236 br = bridge_lookup(argv[1]);
3238 unixctl_command_reply_error(conn, "Unknown bridge");
3241 ofproto_reconnect_controllers(br->ofproto);
3243 HMAP_FOR_EACH (br, node, &all_bridges) {
3244 ofproto_reconnect_controllers(br->ofproto);
3247 unixctl_command_reply(conn, NULL);
3251 bridge_get_controllers(const struct bridge *br,
3252 struct ovsrec_controller ***controllersp)
3254 struct ovsrec_controller **controllers;
3255 size_t n_controllers;
3257 controllers = br->cfg->controller;
3258 n_controllers = br->cfg->n_controller;
3260 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3266 *controllersp = controllers;
3268 return n_controllers;
3272 bridge_collect_wanted_ports(struct bridge *br,
3273 const unsigned long int *splinter_vlans,
3274 struct shash *wanted_ports)
3278 shash_init(wanted_ports);
3280 for (i = 0; i < br->cfg->n_ports; i++) {
3281 const char *name = br->cfg->ports[i]->name;
3282 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3283 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3287 if (bridge_get_controllers(br, NULL)
3288 && !shash_find(wanted_ports, br->name)) {
3289 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3290 br->name, br->name);
3292 ovsrec_interface_init(&br->synth_local_iface);
3293 ovsrec_port_init(&br->synth_local_port);
3295 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3296 br->synth_local_port.n_interfaces = 1;
3297 br->synth_local_port.name = br->name;
3299 br->synth_local_iface.name = br->name;
3300 br->synth_local_iface.type = "internal";
3302 br->synth_local_ifacep = &br->synth_local_iface;
3304 shash_add(wanted_ports, br->name, &br->synth_local_port);
3307 if (splinter_vlans) {
3308 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3312 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3313 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3314 * 'br' needs to complete its configuration. */
3316 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3318 struct shash_node *port_node;
3319 struct port *port, *next;
3321 /* Get rid of deleted ports.
3322 * Get rid of deleted interfaces on ports that still exist. */
3323 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3324 port->cfg = shash_find_data(wanted_ports, port->name);
3328 port_del_ifaces(port);
3332 /* Update iface->cfg and iface->type in interfaces that still exist. */
3333 SHASH_FOR_EACH (port_node, wanted_ports) {
3334 const struct ovsrec_port *port = port_node->data;
3337 for (i = 0; i < port->n_interfaces; i++) {
3338 const struct ovsrec_interface *cfg = port->interfaces[i];
3339 struct iface *iface = iface_lookup(br, cfg->name);
3340 const char *type = iface_get_type(cfg, br->cfg);
3345 } else if (!strcmp(type, "null")) {
3346 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3347 " may be removed in February 2013. Please email"
3348 " dev@openvswitch.org with concerns.",
3351 /* We will add new interfaces later. */
3357 /* Initializes 'oc' appropriately as a management service controller for
3360 * The caller must free oc->target when it is no longer needed. */
3362 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3363 struct ofproto_controller *oc)
3365 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3366 oc->max_backoff = 0;
3367 oc->probe_interval = 60;
3368 oc->band = OFPROTO_OUT_OF_BAND;
3370 oc->burst_limit = 0;
3371 oc->enable_async_msgs = true;
3375 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3377 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3378 struct ofproto_controller *oc)
3382 oc->target = c->target;
3383 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3384 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3385 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3386 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3387 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3388 oc->burst_limit = (c->controller_burst_limit
3389 ? *c->controller_burst_limit : 0);
3390 oc->enable_async_msgs = (!c->enable_async_messages
3391 || *c->enable_async_messages);
3392 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3393 if (dscp < 0 || dscp > 63) {
3394 dscp = DSCP_DEFAULT;
3399 /* Configures the IP stack for 'br''s local interface properly according to the
3400 * configuration in 'c'. */
3402 bridge_configure_local_iface_netdev(struct bridge *br,
3403 struct ovsrec_controller *c)
3405 struct netdev *netdev;
3406 struct in_addr mask, gateway;
3408 struct iface *local_iface;
3411 /* If there's no local interface or no IP address, give up. */
3412 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3413 if (!local_iface || !c->local_ip
3414 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3418 /* Bring up the local interface. */
3419 netdev = local_iface->netdev;
3420 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3422 /* Configure the IP address and netmask. */
3423 if (!c->local_netmask
3424 || !inet_pton(AF_INET, c->local_netmask, &mask)
3426 mask.s_addr = guess_netmask(ip.s_addr);
3428 if (!netdev_set_in4(netdev, ip, mask)) {
3429 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3430 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3433 /* Configure the default gateway. */
3434 if (c->local_gateway
3435 && inet_pton(AF_INET, c->local_gateway, &gateway)
3436 && gateway.s_addr) {
3437 if (!netdev_add_router(netdev, gateway)) {
3438 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3439 br->name, IP_ARGS(gateway.s_addr));
3444 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3445 * in either string are treated as equal to any number of slashes in the other,
3446 * e.g. "x///y" is equal to "x/y".
3448 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3449 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3450 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3451 * 'b' against a prefix of 'a'.
3454 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3456 const char *b_start = b;
3458 if (b - b_start >= b_stoplen) {
3460 } else if (*a != *b) {
3462 } else if (*a == '/') {
3463 a += strspn(a, "/");
3464 b += strspn(b, "/");
3465 } else if (*a == '\0') {
3475 bridge_configure_remotes(struct bridge *br,
3476 const struct sockaddr_in *managers, size_t n_managers)
3478 bool disable_in_band;
3480 struct ovsrec_controller **controllers;
3481 size_t n_controllers;
3483 enum ofproto_fail_mode fail_mode;
3485 struct ofproto_controller *ocs;
3489 /* Check if we should disable in-band control on this bridge. */
3490 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3493 /* Set OpenFlow queue ID for in-band control. */
3494 ofproto_set_in_band_queue(br->ofproto,
3495 smap_get_int(&br->cfg->other_config,
3496 "in-band-queue", -1));
3498 if (disable_in_band) {
3499 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3501 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3504 n_controllers = bridge_get_controllers(br, &controllers);
3506 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3509 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3510 for (i = 0; i < n_controllers; i++) {
3511 struct ovsrec_controller *c = controllers[i];
3513 if (!strncmp(c->target, "punix:", 6)
3514 || !strncmp(c->target, "unix:", 5)) {
3515 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3518 if (!strncmp(c->target, "unix:", 5)) {
3519 /* Connect to a listening socket */
3520 whitelist = xasprintf("unix:%s/", ovs_rundir());
3521 if (strchr(c->target, '/') &&
3522 !equal_pathnames(c->target, whitelist,
3523 strlen(whitelist))) {
3524 /* Absolute path specified, but not in ovs_rundir */
3525 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3526 "controller \"%s\" due to possibility for "
3527 "remote exploit. Instead, specify socket "
3528 "in whitelisted \"%s\" or connect to "
3529 "\"unix:%s/%s.mgmt\" (which is always "
3530 "available without special configuration).",
3531 br->name, c->target, whitelist,
3532 ovs_rundir(), br->name);
3537 whitelist = xasprintf("punix:%s/%s.controller",
3538 ovs_rundir(), br->name);
3539 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3540 /* Prevent remote ovsdb-server users from accessing
3541 * arbitrary Unix domain sockets and overwriting arbitrary
3543 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3544 "controller \"%s\" due to possibility of "
3545 "overwriting local files. Instead, specify "
3546 "whitelisted \"%s\" or connect to "
3547 "\"unix:%s/%s.mgmt\" (which is always "
3548 "available without special configuration).",
3549 br->name, c->target, whitelist,
3550 ovs_rundir(), br->name);
3559 bridge_configure_local_iface_netdev(br, c);
3560 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3561 if (disable_in_band) {
3562 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3567 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3568 bridge_get_allowed_versions(br));
3569 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3572 /* Set the fail-mode. */
3573 fail_mode = !br->cfg->fail_mode
3574 || !strcmp(br->cfg->fail_mode, "standalone")
3575 ? OFPROTO_FAIL_STANDALONE
3576 : OFPROTO_FAIL_SECURE;
3577 ofproto_set_fail_mode(br->ofproto, fail_mode);
3579 /* Configure OpenFlow controller connection snooping. */
3580 if (!ofproto_has_snoops(br->ofproto)) {
3584 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3585 ovs_rundir(), br->name));
3586 ofproto_set_snoops(br->ofproto, &snoops);
3587 sset_destroy(&snoops);
3592 bridge_configure_tables(struct bridge *br)
3594 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3598 n_tables = ofproto_get_n_tables(br->ofproto);
3600 for (i = 0; i < n_tables; i++) {
3601 struct ofproto_table_settings s;
3602 bool use_default_prefixes = true;
3605 s.max_flows = UINT_MAX;
3607 s.enable_eviction = false;
3609 s.n_prefix_fields = 0;
3610 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3612 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3613 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3616 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3617 s.max_flows = *cfg->flow_limit;
3620 s.enable_eviction = (cfg->overflow_policy
3621 && !strcmp(cfg->overflow_policy, "evict"));
3622 if (cfg->n_groups) {
3623 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3624 for (k = 0; k < cfg->n_groups; k++) {
3625 const char *string = cfg->groups[k];
3628 msg = mf_parse_subfield__(&s.groups[k], &string);
3630 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3631 "'groups' (%s)", br->name, i, msg);
3633 } else if (*string) {
3634 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3635 "element '%s' contains trailing garbage",
3636 br->name, i, cfg->groups[k]);
3643 /* Prefix lookup fields. */
3644 s.n_prefix_fields = 0;
3645 for (k = 0; k < cfg->n_prefixes; k++) {
3646 const char *name = cfg->prefixes[k];
3647 const struct mf_field *mf;
3649 if (strcmp(name, "none") == 0) {
3650 use_default_prefixes = false;
3651 s.n_prefix_fields = 0;
3654 mf = mf_from_name(name);
3656 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3660 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3661 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3662 "%s", br->name, name);
3665 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3666 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3667 "field not used: %s", br->name, name);
3670 use_default_prefixes = false;
3671 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3674 if (use_default_prefixes) {
3675 /* Use default values. */
3676 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3677 memcpy(s.prefix_fields, default_prefix_fields,
3678 sizeof default_prefix_fields);
3681 struct ds ds = DS_EMPTY_INITIALIZER;
3682 for (k = 0; k < s.n_prefix_fields; k++) {
3684 ds_put_char(&ds, ',');
3686 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3688 if (s.n_prefix_fields == 0) {
3689 ds_put_cstr(&ds, "none");
3691 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3692 br->name, i, ds_cstr(&ds));
3696 ofproto_configure_table(br->ofproto, i, &s);
3700 for (; j < br->cfg->n_flow_tables; j++) {
3701 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3702 "%"PRId64" not supported by this datapath", br->name,
3703 br->cfg->key_flow_tables[j]);
3708 bridge_configure_dp_desc(struct bridge *br)
3710 ofproto_set_dp_desc(br->ofproto,
3711 smap_get(&br->cfg->other_config, "dp-desc"));
3714 static struct aa_mapping *
3715 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3717 struct aa_mapping *m;
3719 HMAP_FOR_EACH_IN_BUCKET (m,
3721 hash_bytes(&isid, sizeof isid, 0),
3723 if (isid == m->isid) {
3730 static struct aa_mapping *
3731 bridge_aa_mapping_create(struct bridge *br,
3735 struct aa_mapping *m;
3737 m = xzalloc(sizeof *m);
3741 m->br_name = xstrdup(br->name);
3742 hmap_insert(&br->mappings,
3744 hash_bytes(&isid, sizeof isid, 0));
3750 bridge_aa_mapping_destroy(struct aa_mapping *m)
3753 struct bridge *br = m->bridge;
3756 ofproto_aa_mapping_unregister(br->ofproto, m);
3759 hmap_remove(&br->mappings, &m->hmap_node);
3768 bridge_aa_mapping_configure(struct aa_mapping *m)
3770 struct aa_mapping_settings s;
3776 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3782 bridge_configure_aa(struct bridge *br)
3784 const struct ovsdb_datum *mc;
3785 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3786 struct aa_settings aa_s;
3787 struct aa_mapping *m, *next;
3791 ofproto_set_aa(br->ofproto, NULL, NULL);
3795 memset(&aa_s, 0, sizeof aa_s);
3796 aa_s.system_description = auto_attach->system_description;
3797 aa_s.system_name = auto_attach->system_name;
3798 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3800 mc = ovsrec_autoattach_get_mappings(auto_attach,
3802 OVSDB_TYPE_INTEGER);
3803 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3804 union ovsdb_atom atom;
3806 atom.integer = m->isid;
3807 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3808 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3810 bridge_aa_mapping_destroy(m);
3814 /* Add new mappings and reconfigure existing ones. */
3815 for (i = 0; i < auto_attach->n_mappings; ++i) {
3816 struct aa_mapping *m =
3817 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3820 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3821 auto_attach->key_mappings[i],
3822 auto_attach->value_mappings[i]);
3823 m = bridge_aa_mapping_create(br,
3824 auto_attach->key_mappings[i],
3825 auto_attach->value_mappings[i]);
3827 if (!bridge_aa_mapping_configure(m)) {
3828 bridge_aa_mapping_destroy(m);
3835 bridge_aa_need_refresh(struct bridge *br)
3837 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3841 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3843 int64_t *trunks = NULL;
3845 bool found = false, reconfigure = false;
3847 for (i = 0; i < port->cfg->n_trunks; i++) {
3848 if (port->cfg->trunks[i] == m->vlan) {
3855 case BRIDGE_AA_VLAN_OPER_ADD:
3857 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3859 for (i = 0; i < port->cfg->n_trunks; i++) {
3860 trunks[i] = port->cfg->trunks[i];
3862 trunks[i++] = m->vlan;
3868 case BRIDGE_AA_VLAN_OPER_REMOVE:
3872 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3874 for (i = 0; i < port->cfg->n_trunks; i++) {
3875 if (port->cfg->trunks[i] != m->vlan) {
3876 trunks[j++] = port->cfg->trunks[i];
3885 case BRIDGE_AA_VLAN_OPER_UNDEF:
3887 VLOG_WARN("unrecognized operation %u", m->oper);
3892 /* VLAN switching under trunk mode cause the trunk port to switch all
3893 * VLANs, see ovs-vswitchd.conf.db
3896 static char *vlan_mode_access = "access";
3897 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3901 static char *vlan_mode_trunk = "trunk";
3902 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3905 ovsrec_port_set_trunks(port->cfg, trunks, i);
3907 /* Force reconfigure of the port. */
3908 port_configure(port);
3913 bridge_aa_refresh_queued(struct bridge *br)
3915 struct ovs_list *list = xmalloc(sizeof *list);
3916 struct bridge_aa_vlan *node, *next;
3919 ofproto_aa_vlan_get_queued(br->ofproto, list);
3921 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3924 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3927 port = port_lookup(br, node->port_name);
3929 bridge_aa_update_trunks(port, node);
3932 list_remove(&node->list_node);
3933 free(node->port_name);
3941 /* Port functions. */
3943 static struct port *
3944 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3948 port = xzalloc(sizeof *port);
3950 port->name = xstrdup(cfg->name);
3952 list_init(&port->ifaces);
3954 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3958 /* Deletes interfaces from 'port' that are no longer configured for it. */
3960 port_del_ifaces(struct port *port)
3962 struct iface *iface, *next;
3963 struct sset new_ifaces;
3966 /* Collect list of new interfaces. */
3967 sset_init(&new_ifaces);
3968 for (i = 0; i < port->cfg->n_interfaces; i++) {
3969 const char *name = port->cfg->interfaces[i]->name;
3970 const char *type = port->cfg->interfaces[i]->type;
3971 if (strcmp(type, "null")) {
3972 sset_add(&new_ifaces, name);
3976 /* Get rid of deleted interfaces. */
3977 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3978 if (!sset_contains(&new_ifaces, iface->name)) {
3979 iface_destroy(iface);
3983 sset_destroy(&new_ifaces);
3987 port_destroy(struct port *port)
3990 struct bridge *br = port->bridge;
3991 struct iface *iface, *next;
3994 ofproto_bundle_unregister(br->ofproto, port);
3997 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3998 iface_destroy__(iface);
4001 hmap_remove(&br->ports, &port->hmap_node);
4007 static struct port *
4008 port_lookup(const struct bridge *br, const char *name)
4012 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4014 if (!strcmp(port->name, name)) {
4022 enable_lacp(struct port *port, bool *activep)
4024 if (!port->cfg->lacp) {
4025 /* XXX when LACP implementation has been sufficiently tested, enable by
4026 * default and make active on bonded ports. */
4028 } else if (!strcmp(port->cfg->lacp, "off")) {
4030 } else if (!strcmp(port->cfg->lacp, "active")) {
4033 } else if (!strcmp(port->cfg->lacp, "passive")) {
4037 VLOG_WARN("port %s: unknown LACP mode %s",
4038 port->name, port->cfg->lacp);
4043 static struct lacp_settings *
4044 port_configure_lacp(struct port *port, struct lacp_settings *s)
4046 const char *lacp_time, *system_id;
4049 if (!enable_lacp(port, &s->active)) {
4053 s->name = port->name;
4055 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4057 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4058 ETH_ADDR_SCAN_ARGS(s->id))) {
4059 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4060 " address.", port->name, system_id);
4064 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4067 if (eth_addr_is_zero(s->id)) {
4068 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4072 /* Prefer bondable links if unspecified. */
4073 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4075 s->priority = (priority > 0 && priority <= UINT16_MAX
4077 : UINT16_MAX - !list_is_short(&port->ifaces));
4079 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4080 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4082 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4083 "lacp-fallback-ab", false);
4089 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4091 int priority, portid, key;
4093 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4094 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4096 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4098 if (portid <= 0 || portid > UINT16_MAX) {
4099 portid = ofp_to_u16(iface->ofp_port);
4102 if (priority <= 0 || priority > UINT16_MAX) {
4103 priority = UINT16_MAX;
4106 if (key < 0 || key > UINT16_MAX) {
4110 s->name = iface->name;
4112 s->priority = priority;
4117 port_configure_bond(struct port *port, struct bond_settings *s)
4119 const char *detect_s;
4120 struct iface *iface;
4122 int miimon_interval;
4124 s->name = port->name;
4126 if (port->cfg->bond_mode) {
4127 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4128 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4129 port->name, port->cfg->bond_mode,
4130 bond_mode_to_string(s->balance));
4133 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4135 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4136 * active-backup. At some point we should remove this warning. */
4137 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4138 " in previous versions, the default bond_mode was"
4139 " balance-slb", port->name,
4140 bond_mode_to_string(s->balance));
4142 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4143 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4144 "please use another bond type or disable flood_vlans",
4148 miimon_interval = smap_get_int(&port->cfg->other_config,
4149 "bond-miimon-interval", 0);
4150 if (miimon_interval <= 0) {
4151 miimon_interval = 200;
4154 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4155 if (!detect_s || !strcmp(detect_s, "carrier")) {
4156 miimon_interval = 0;
4157 } else if (strcmp(detect_s, "miimon")) {
4158 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4159 "defaulting to carrier", port->name, detect_s);
4160 miimon_interval = 0;
4163 s->up_delay = MAX(0, port->cfg->bond_updelay);
4164 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4165 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4166 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4167 "bond-rebalance-interval", 10000);
4168 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4169 s->rebalance_interval = 1000;
4172 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4173 "lacp-fallback-ab", false);
4175 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4176 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4179 mac_s = port->cfg->bond_active_slave;
4180 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4181 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4182 /* OVSDB did not store the last active interface */
4183 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4187 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4188 * instead of obtaining it from the database. */
4190 port_is_synthetic(const struct port *port)
4192 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4195 /* Interface functions. */
4198 iface_is_internal(const struct ovsrec_interface *iface,
4199 const struct ovsrec_bridge *br)
4201 /* The local port and "internal" ports are always "internal". */
4202 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4205 /* Returns the correct network device type for interface 'iface' in bridge
4208 iface_get_type(const struct ovsrec_interface *iface,
4209 const struct ovsrec_bridge *br)
4213 /* The local port always has type "internal". Other ports take
4214 * their type from the database and default to "system" if none is
4216 if (iface_is_internal(iface, br)) {
4219 type = iface->type[0] ? iface->type : "system";
4222 return ofproto_port_open_type(br->datapath_type, type);
4226 iface_destroy__(struct iface *iface)
4229 struct port *port = iface->port;
4230 struct bridge *br = port->bridge;
4232 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4233 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4236 if (iface->ofp_port != OFPP_NONE) {
4237 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4240 list_remove(&iface->port_elem);
4241 hmap_remove(&br->iface_by_name, &iface->name_node);
4243 /* The user is changing configuration here, so netdev_remove needs to be
4244 * used as opposed to netdev_close */
4245 netdev_remove(iface->netdev);
4253 iface_destroy(struct iface *iface)
4256 struct port *port = iface->port;
4258 iface_destroy__(iface);
4259 if (list_is_empty(&port->ifaces)) {
4265 static struct iface *
4266 iface_lookup(const struct bridge *br, const char *name)
4268 struct iface *iface;
4270 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4271 &br->iface_by_name) {
4272 if (!strcmp(iface->name, name)) {
4280 static struct iface *
4281 iface_find(const char *name)
4283 const struct bridge *br;
4285 HMAP_FOR_EACH (br, node, &all_bridges) {
4286 struct iface *iface = iface_lookup(br, name);
4295 static struct iface *
4296 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4298 struct iface *iface;
4300 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4302 if (iface->ofp_port == ofp_port) {
4309 /* Set Ethernet address of 'iface', if one is specified in the configuration
4312 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4314 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4315 struct iface *hw_addr_iface;
4317 if (strcmp(iface->type, "internal")) {
4321 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4323 } else if (port->cfg->fake_bridge) {
4324 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4325 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4330 if (iface->ofp_port == OFPP_LOCAL) {
4331 VLOG_ERR("interface %s: ignoring mac in Interface record "
4332 "(use Bridge record to set local port's mac)",
4334 } else if (eth_addr_is_multicast(mac)) {
4335 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4338 int error = netdev_set_etheraddr(iface->netdev, mac);
4340 VLOG_ERR("interface %s: setting MAC failed (%s)",
4341 iface->name, ovs_strerror(error));
4347 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4349 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4351 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4352 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4353 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4357 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4358 * sets the "ofport" field to -1.
4360 * This is appropriate when 'if_cfg''s interface cannot be created or is
4361 * otherwise invalid. */
4363 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4365 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4366 iface_set_ofport(if_cfg, OFPP_NONE);
4367 ovsrec_interface_set_error(if_cfg, errp);
4368 ovsrec_interface_set_status(if_cfg, NULL);
4369 ovsrec_interface_set_admin_state(if_cfg, NULL);
4370 ovsrec_interface_set_duplex(if_cfg, NULL);
4371 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4372 ovsrec_interface_set_link_state(if_cfg, NULL);
4373 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4374 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4375 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4376 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4377 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4378 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4379 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4380 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4385 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4387 union ovsdb_atom atom;
4389 atom.integer = target;
4390 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4394 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4396 struct ofpbuf queues_buf;
4398 ofpbuf_init(&queues_buf, 0);
4400 if (!qos || qos->type[0] == '\0') {
4401 netdev_set_qos(iface->netdev, NULL, NULL);
4403 const struct ovsdb_datum *queues;
4404 struct netdev_queue_dump dump;
4405 unsigned int queue_id;
4406 struct smap details;
4410 /* Configure top-level Qos for 'iface'. */
4411 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4413 /* Deconfigure queues that were deleted. */
4414 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4416 smap_init(&details);
4417 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4418 if (!queue_ids_include(queues, queue_id)) {
4419 netdev_delete_queue(iface->netdev, queue_id);
4422 smap_destroy(&details);
4424 /* Configure queues for 'iface'. */
4426 for (i = 0; i < qos->n_queues; i++) {
4427 const struct ovsrec_queue *queue = qos->value_queues[i];
4428 unsigned int queue_id = qos->key_queues[i];
4430 if (queue_id == 0) {
4434 if (queue->n_dscp == 1) {
4435 struct ofproto_port_queue *port_queue;
4437 port_queue = ofpbuf_put_uninit(&queues_buf,
4438 sizeof *port_queue);
4439 port_queue->queue = queue_id;
4440 port_queue->dscp = queue->dscp[0];
4443 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4446 struct smap details;
4448 smap_init(&details);
4449 netdev_set_queue(iface->netdev, 0, &details);
4450 smap_destroy(&details);
4454 if (iface->ofp_port != OFPP_NONE) {
4455 const struct ofproto_port_queue *port_queues = queues_buf.data;
4456 size_t n_queues = queues_buf.size / sizeof *port_queues;
4458 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4459 port_queues, n_queues);
4462 netdev_set_policing(iface->netdev,
4463 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4464 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4466 ofpbuf_uninit(&queues_buf);
4470 iface_configure_cfm(struct iface *iface)
4472 const struct ovsrec_interface *cfg = iface->cfg;
4473 const char *opstate_str;
4474 const char *cfm_ccm_vlan;
4475 struct cfm_settings s;
4476 struct smap netdev_args;
4478 if (!cfg->n_cfm_mpid) {
4479 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4483 s.check_tnl_key = false;
4484 smap_init(&netdev_args);
4485 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4486 const char *key = smap_get(&netdev_args, "key");
4487 const char *in_key = smap_get(&netdev_args, "in_key");
4489 s.check_tnl_key = (key && !strcmp(key, "flow"))
4490 || (in_key && !strcmp(in_key, "flow"));
4492 smap_destroy(&netdev_args);
4494 s.mpid = *cfg->cfm_mpid;
4495 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4496 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4497 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4499 if (s.interval <= 0) {
4503 if (!cfm_ccm_vlan) {
4505 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4506 s.ccm_vlan = CFM_RANDOM_VLAN;
4508 s.ccm_vlan = atoi(cfm_ccm_vlan);
4509 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4514 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4516 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4518 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4519 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4521 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4524 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4525 * instead of obtaining it from the database. */
4527 iface_is_synthetic(const struct iface *iface)
4529 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4533 iface_validate_ofport__(size_t n, int64_t *ofport)
4535 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4536 ? u16_to_ofp(*ofport)
4541 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4543 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4547 iface_pick_ofport(const struct ovsrec_interface *cfg)
4549 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4550 return (requested_ofport != OFPP_NONE
4552 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4555 /* Port mirroring. */
4557 static struct mirror *
4558 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4562 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4563 if (uuid_equals(uuid, &m->uuid)) {
4571 bridge_configure_mirrors(struct bridge *br)
4573 const struct ovsdb_datum *mc;
4574 unsigned long *flood_vlans;
4575 struct mirror *m, *next;
4578 /* Get rid of deleted mirrors. */
4579 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4580 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4581 union ovsdb_atom atom;
4583 atom.uuid = m->uuid;
4584 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4589 /* Add new mirrors and reconfigure existing ones. */
4590 for (i = 0; i < br->cfg->n_mirrors; i++) {
4591 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4592 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4594 m = mirror_create(br, cfg);
4597 if (!mirror_configure(m)) {
4602 /* Update flooded vlans (for RSPAN). */
4603 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4604 br->cfg->n_flood_vlans);
4605 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4606 bitmap_free(flood_vlans);
4609 static struct mirror *
4610 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4614 m = xzalloc(sizeof *m);
4615 m->uuid = cfg->header_.uuid;
4616 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4618 m->name = xstrdup(cfg->name);
4624 mirror_destroy(struct mirror *m)
4627 struct bridge *br = m->bridge;
4630 ofproto_mirror_unregister(br->ofproto, m);
4633 hmap_remove(&br->mirrors, &m->hmap_node);
4640 mirror_collect_ports(struct mirror *m,
4641 struct ovsrec_port **in_ports, int n_in_ports,
4642 void ***out_portsp, size_t *n_out_portsp)
4644 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4645 size_t n_out_ports = 0;
4648 for (i = 0; i < n_in_ports; i++) {
4649 const char *name = in_ports[i]->name;
4650 struct port *port = port_lookup(m->bridge, name);
4652 out_ports[n_out_ports++] = port;
4654 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4655 "port %s", m->bridge->name, m->name, name);
4658 *out_portsp = out_ports;
4659 *n_out_portsp = n_out_ports;
4663 mirror_configure(struct mirror *m)
4665 const struct ovsrec_mirror *cfg = m->cfg;
4666 struct ofproto_mirror_settings s;
4669 if (strcmp(cfg->name, m->name)) {
4671 m->name = xstrdup(cfg->name);
4675 /* Get output port or VLAN. */
4676 if (cfg->output_port) {
4677 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4678 if (!s.out_bundle) {
4679 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4680 m->bridge->name, m->name);
4683 s.out_vlan = UINT16_MAX;
4685 if (cfg->output_vlan) {
4686 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4687 "output vlan; ignoring output vlan",
4688 m->bridge->name, m->name);
4690 } else if (cfg->output_vlan) {
4691 /* The database should prevent invalid VLAN values. */
4692 s.out_bundle = NULL;
4693 s.out_vlan = *cfg->output_vlan;
4695 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4696 m->bridge->name, m->name);
4700 /* Get port selection. */
4701 if (cfg->select_all) {
4702 size_t n_ports = hmap_count(&m->bridge->ports);
4703 void **ports = xmalloc(n_ports * sizeof *ports);
4708 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4718 /* Get ports, dropping ports that don't exist.
4719 * The IDL ensures that there are no duplicates. */
4720 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4721 &s.srcs, &s.n_srcs);
4722 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4723 &s.dsts, &s.n_dsts);
4726 /* Get VLAN selection. */
4727 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4730 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4733 if (s.srcs != s.dsts) {
4742 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4744 * This is deprecated. It is only for compatibility with broken device drivers
4745 * in old versions of Linux that do not properly support VLANs when VLAN
4746 * devices are not used. When broken device drivers are no longer in
4747 * widespread use, we will delete these interfaces. */
4749 static struct ovsrec_port **recs;
4750 static size_t n_recs, allocated_recs;
4752 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4753 * splinters are reconfigured. */
4755 register_rec(struct ovsrec_port *rec)
4757 if (n_recs >= allocated_recs) {
4758 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4760 recs[n_recs++] = rec;
4763 /* Frees all of the ports registered with register_reg(). */
4765 free_registered_recs(void)
4769 for (i = 0; i < n_recs; i++) {
4770 struct ovsrec_port *port = recs[i];
4773 for (j = 0; j < port->n_interfaces; j++) {
4774 struct ovsrec_interface *iface = port->interfaces[j];
4779 smap_destroy(&port->other_config);
4780 free(port->interfaces);
4788 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4791 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4793 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4797 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4800 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4801 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4802 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4805 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4806 * use, returns NULL.
4808 * Updates 'vlan_splinters_enabled_anywhere'. */
4809 static unsigned long int *
4810 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4812 unsigned long int *splinter_vlans;
4813 struct sset splinter_ifaces;
4814 const char *real_dev_name;
4815 struct shash *real_devs;
4816 struct shash_node *node;
4820 /* Free space allocated for synthesized ports and interfaces, since we're
4821 * in the process of reconstructing all of them. */
4822 free_registered_recs();
4824 splinter_vlans = bitmap_allocate(4096);
4825 sset_init(&splinter_ifaces);
4826 vlan_splinters_enabled_anywhere = false;
4827 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4828 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4831 for (j = 0; j < br_cfg->n_ports; j++) {
4832 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4835 for (k = 0; k < port_cfg->n_interfaces; k++) {
4836 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4838 if (vlan_splinters_is_enabled(iface_cfg)) {
4839 vlan_splinters_enabled_anywhere = true;
4840 sset_add(&splinter_ifaces, iface_cfg->name);
4841 vlan_bitmap_from_array__(port_cfg->trunks,
4847 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4848 bitmap_set1(splinter_vlans, *port_cfg->tag);
4853 if (!vlan_splinters_enabled_anywhere) {
4854 free(splinter_vlans);
4855 sset_destroy(&splinter_ifaces);
4859 HMAP_FOR_EACH (br, node, &all_bridges) {
4861 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4865 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4866 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4867 * device to be created for it. */
4868 bitmap_set0(splinter_vlans, 0);
4869 bitmap_set0(splinter_vlans, 4095);
4871 /* Delete all VLAN devices that we don't need. */
4873 real_devs = vlandev_get_real_devs();
4874 SHASH_FOR_EACH (node, real_devs) {
4875 const struct vlan_real_dev *real_dev = node->data;
4876 const struct vlan_dev *vlan_dev;
4877 bool real_dev_has_splinters;
4879 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4881 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4882 if (!real_dev_has_splinters
4883 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4884 struct netdev *netdev;
4886 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4887 if (!netdev_get_in4(netdev, NULL, NULL) ||
4888 !netdev_get_in6(netdev, NULL)) {
4889 /* It has an IP address configured, so we don't own
4890 * it. Don't delete it. */
4892 vlandev_del(vlan_dev->name);
4894 netdev_close(netdev);
4901 /* Add all VLAN devices that we need. */
4902 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4905 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4906 if (!vlandev_get_name(real_dev_name, vid)) {
4907 vlandev_add(real_dev_name, vid);
4914 sset_destroy(&splinter_ifaces);
4916 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4917 free(splinter_vlans);
4920 return splinter_vlans;
4923 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4926 configure_splinter_port(struct port *port)
4928 struct ofproto *ofproto = port->bridge->ofproto;
4929 ofp_port_t realdev_ofp_port;
4930 const char *realdev_name;
4931 struct iface *vlandev, *realdev;
4933 ofproto_bundle_unregister(port->bridge->ofproto, port);
4935 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4938 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4939 realdev = iface_lookup(port->bridge, realdev_name);
4940 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4942 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4946 static struct ovsrec_port *
4947 synthesize_splinter_port(const char *real_dev_name,
4948 const char *vlan_dev_name, int vid)
4950 struct ovsrec_interface *iface;
4951 struct ovsrec_port *port;
4953 iface = xmalloc(sizeof *iface);
4954 ovsrec_interface_init(iface);
4955 iface->name = xstrdup(vlan_dev_name);
4956 iface->type = "system";
4958 port = xmalloc(sizeof *port);
4959 ovsrec_port_init(port);
4960 port->interfaces = xmemdup(&iface, sizeof iface);
4961 port->n_interfaces = 1;
4962 port->name = xstrdup(vlan_dev_name);
4963 port->vlan_mode = "splinter";
4964 port->tag = xmalloc(sizeof *port->tag);
4967 smap_add(&port->other_config, "realdev", real_dev_name);
4973 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4974 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4975 * 1-bit in the 'splinter_vlans' bitmap. */
4977 add_vlan_splinter_ports(struct bridge *br,
4978 const unsigned long int *splinter_vlans,
4979 struct shash *ports)
4983 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4984 * we're modifying 'ports'. */
4985 for (i = 0; i < br->cfg->n_ports; i++) {
4986 const char *name = br->cfg->ports[i]->name;
4987 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4990 for (j = 0; j < port_cfg->n_interfaces; j++) {
4991 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4993 if (vlan_splinters_is_enabled(iface_cfg)) {
4994 const char *real_dev_name;
4997 real_dev_name = iface_cfg->name;
4998 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4999 const char *vlan_dev_name;
5001 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5003 && !shash_find(ports, vlan_dev_name)) {
5004 shash_add(ports, vlan_dev_name,
5005 synthesize_splinter_port(
5006 real_dev_name, vlan_dev_name, vid));
5015 mirror_refresh_stats(struct mirror *m)
5017 struct ofproto *ofproto = m->bridge->ofproto;
5018 uint64_t tx_packets, tx_bytes;
5019 const char *keys[2];
5021 size_t stat_cnt = 0;
5023 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5024 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5028 if (tx_packets != UINT64_MAX) {
5029 keys[stat_cnt] = "tx_packets";
5030 values[stat_cnt] = tx_packets;
5033 if (tx_bytes != UINT64_MAX) {
5034 keys[stat_cnt] = "tx_bytes";
5035 values[stat_cnt] = tx_bytes;
5039 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5043 * Add registered netdev and dpif types to ovsdb to allow external
5044 * applications to query the capabilities of the Open vSwitch instance
5045 * running on the node.
5048 discover_types(const struct ovsrec_open_vswitch *cfg)
5052 /* Datapath types. */
5054 dp_enumerate_types(&types);
5055 const char **datapath_types = sset_array(&types);
5056 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5057 sset_count(&types));
5058 free(datapath_types);
5059 sset_destroy(&types);
5063 netdev_enumerate_types(&types);
5064 const char **iface_types = sset_array(&types);
5065 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5067 sset_destroy(&types);