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"
51 #include "if-notifier.h"
56 #include "socket-util.h"
58 #include "stream-ssl.h"
60 #include "system-stats.h"
65 #include "lib/vswitch-idl.h"
66 #include "xenserver.h"
67 #include "openvswitch/vlog.h"
68 #include "sflow_api.h"
69 #include "vlan-bitmap.h"
72 VLOG_DEFINE_THIS_MODULE(bridge);
74 COVERAGE_DEFINE(bridge_reconfigure);
77 /* These members are always valid.
79 * They are immutable: they never change between iface_create() and
81 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
86 struct netdev *netdev; /* Network device. */
87 ofp_port_t ofp_port; /* OpenFlow port number. */
90 /* These members are valid only within bridge_reconfigure(). */
91 const char *type; /* Usually same as cfg->type. */
92 const struct ovsrec_interface *cfg;
96 struct uuid uuid; /* UUID of this "mirror" record in database. */
97 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
98 struct bridge *bridge;
100 const struct ovsrec_mirror *cfg;
104 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
105 struct bridge *bridge;
108 const struct ovsrec_port *cfg;
110 /* An ordinary bridge port has 1 interface.
111 * A bridge port for bonding has at least 2 interfaces. */
112 struct ovs_list ifaces; /* List of "struct iface"s. */
116 struct hmap_node node; /* In 'all_bridges'. */
117 char *name; /* User-specified arbitrary name. */
118 char *type; /* Datapath type. */
119 struct eth_addr ea; /* Bridge Ethernet Address. */
120 struct eth_addr default_ea; /* Default MAC. */
121 const struct ovsrec_bridge *cfg;
123 /* OpenFlow switch processing. */
124 struct ofproto *ofproto; /* OpenFlow switch. */
127 struct hmap ports; /* "struct port"s indexed by name. */
128 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
129 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
131 /* Port mirroring. */
132 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
135 struct hmap mappings; /* "struct" indexed by UUID */
137 /* Used during reconfiguration. */
138 struct shash wanted_ports;
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
147 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
148 struct bridge *bridge;
154 /* All bridges, indexed by name. */
155 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
157 /* OVSDB IDL used to obtain configuration. */
158 static struct ovsdb_idl *idl;
160 /* We want to complete daemonization, fully detaching from our parent process,
161 * only after we have completed our initial configuration, committed our state
162 * to the database, and received confirmation back from the database server
163 * that it applied the commit. This allows our parent process to know that,
164 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
165 * to have already been filled in. (It is only "very likely" rather than
166 * certain because there is always a slim possibility that the transaction will
167 * fail or that some other client has added new bridges, ports, etc. while
168 * ovs-vswitchd was configuring using an old configuration.)
170 * We only need to do this once for our initial configuration at startup, so
171 * 'initial_config_done' tracks whether we've already done it. While we are
172 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
173 * itself and is otherwise NULL. */
174 static bool initial_config_done;
175 static struct ovsdb_idl_txn *daemonize_txn;
177 /* Most recently processed IDL sequence number. */
178 static unsigned int idl_seqno;
180 /* Track changes to port connectivity. */
181 static uint64_t connectivity_seqno = LLONG_MIN;
183 /* Status update to database.
185 * Some information in the database must be kept as up-to-date as possible to
186 * allow controllers to respond rapidly to network outages. Those status are
187 * updated via the 'status_txn'.
189 * We use the global connectivity sequence number to detect the status change.
190 * Also, to prevent the status update from sending too much to the database,
191 * we check the return status of each update transaction and do not start new
192 * update if the previous transaction status is 'TXN_INCOMPLETE'.
194 * 'statux_txn' is NULL if there is no ongoing status update.
196 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
197 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
198 * which will cause the main thread wake up soon and retry the status update.
200 static struct ovsdb_idl_txn *status_txn;
201 static bool status_txn_try_again;
203 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
204 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
205 #define STATUS_CHECK_AGAIN_MSEC 100
207 /* Statistics update to database. */
208 static struct ovsdb_idl_txn *stats_txn;
210 /* Each time this timer expires, the bridge fetches interface and mirror
211 * statistics and pushes them into the database. */
212 static int stats_timer_interval;
213 static long long int stats_timer = LLONG_MIN;
215 /* Each time this timer expires, the bridge fetches the list of port/VLAN
216 * membership that has been modified by the AA.
218 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
219 static long long int aa_refresh_timer = LLONG_MIN;
221 /* Whenever system interfaces are added, removed or change state, the bridge
222 * will be reconfigured.
224 static struct if_notifier *ifnotifier;
225 static bool ifaces_changed = false;
227 static void add_del_bridges(const struct ovsrec_open_vswitch *);
228 static void bridge_run__(void);
229 static void bridge_create(const struct ovsrec_bridge *);
230 static void bridge_destroy(struct bridge *);
231 static struct bridge *bridge_lookup(const char *name);
232 static unixctl_cb_func bridge_unixctl_dump_flows;
233 static unixctl_cb_func bridge_unixctl_reconnect;
234 static size_t bridge_get_controllers(const struct bridge *br,
235 struct ovsrec_controller ***controllersp);
236 static void bridge_collect_wanted_ports(struct bridge *,
237 const unsigned long *splinter_vlans,
238 struct shash *wanted_ports);
239 static void bridge_delete_ofprotos(void);
240 static void bridge_delete_or_reconfigure_ports(struct bridge *);
241 static void bridge_del_ports(struct bridge *,
242 const struct shash *wanted_ports);
243 static void bridge_add_ports(struct bridge *,
244 const struct shash *wanted_ports);
246 static void bridge_configure_datapath_id(struct bridge *);
247 static void bridge_configure_netflow(struct bridge *);
248 static void bridge_configure_forward_bpdu(struct bridge *);
249 static void bridge_configure_mac_table(struct bridge *);
250 static void bridge_configure_mcast_snooping(struct bridge *);
251 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
252 static void bridge_configure_ipfix(struct bridge *);
253 static void bridge_configure_spanning_tree(struct bridge *);
254 static void bridge_configure_tables(struct bridge *);
255 static void bridge_configure_dp_desc(struct bridge *);
256 static void bridge_configure_aa(struct bridge *);
257 static void bridge_aa_refresh_queued(struct bridge *);
258 static bool bridge_aa_need_refresh(struct bridge *);
259 static void bridge_configure_remotes(struct bridge *,
260 const struct sockaddr_in *managers,
262 static void bridge_pick_local_hw_addr(struct bridge *, struct eth_addr *ea,
263 struct iface **hw_addr_iface);
264 static uint64_t bridge_pick_datapath_id(struct bridge *,
265 const struct eth_addr bridge_ea,
266 struct iface *hw_addr_iface);
267 static uint64_t dpid_from_hash(const void *, size_t nbytes);
268 static bool bridge_has_bond_fake_iface(const struct bridge *,
270 static bool port_is_bond_fake_iface(const struct port *);
272 static unixctl_cb_func qos_unixctl_show;
274 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
275 static void port_del_ifaces(struct port *);
276 static void port_destroy(struct port *);
277 static struct port *port_lookup(const struct bridge *, const char *name);
278 static void port_configure(struct port *);
279 static struct lacp_settings *port_configure_lacp(struct port *,
280 struct lacp_settings *);
281 static void port_configure_bond(struct port *, struct bond_settings *);
282 static bool port_is_synthetic(const struct port *);
284 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
285 static void run_system_stats(void);
287 static void bridge_configure_mirrors(struct bridge *);
288 static struct mirror *mirror_create(struct bridge *,
289 const struct ovsrec_mirror *);
290 static void mirror_destroy(struct mirror *);
291 static bool mirror_configure(struct mirror *);
292 static void mirror_refresh_stats(struct mirror *);
294 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
295 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
296 const struct ovsrec_port *);
297 static bool iface_is_internal(const struct ovsrec_interface *iface,
298 const struct ovsrec_bridge *br);
299 static const char *iface_get_type(const struct ovsrec_interface *,
300 const struct ovsrec_bridge *);
301 static void iface_destroy(struct iface *);
302 static void iface_destroy__(struct iface *);
303 static struct iface *iface_lookup(const struct bridge *, const char *name);
304 static struct iface *iface_find(const char *name);
305 static struct iface *iface_from_ofp_port(const struct bridge *,
306 ofp_port_t ofp_port);
307 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
308 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
309 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
310 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
311 static void iface_configure_cfm(struct iface *);
312 static void iface_refresh_cfm_stats(struct iface *);
313 static void iface_refresh_stats(struct iface *);
314 static void iface_refresh_netdev_status(struct iface *);
315 static void iface_refresh_ofproto_status(struct iface *);
316 static bool iface_is_synthetic(const struct iface *);
317 static ofp_port_t iface_get_requested_ofp_port(
318 const struct ovsrec_interface *);
319 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
322 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
324 * This is deprecated. It is only for compatibility with broken device drivers
325 * in old versions of Linux that do not properly support VLANs when VLAN
326 * devices are not used. When broken device drivers are no longer in
327 * widespread use, we will delete these interfaces. */
329 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
330 static bool vlan_splinters_enabled_anywhere;
332 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
333 static unsigned long int *collect_splinter_vlans(
334 const struct ovsrec_open_vswitch *);
335 static void configure_splinter_port(struct port *);
336 static void add_vlan_splinter_ports(struct bridge *,
337 const unsigned long int *splinter_vlans,
338 struct shash *ports);
340 static void discover_types(const struct ovsrec_open_vswitch *cfg);
343 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
345 struct shash iface_hints;
346 static bool initialized = false;
353 shash_init(&iface_hints);
356 for (i = 0; i < cfg->n_bridges; i++) {
357 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
360 for (j = 0; j < br_cfg->n_ports; j++) {
361 struct ovsrec_port *port_cfg = br_cfg->ports[j];
364 for (k = 0; k < port_cfg->n_interfaces; k++) {
365 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
366 struct iface_hint *iface_hint;
368 iface_hint = xmalloc(sizeof *iface_hint);
369 iface_hint->br_name = br_cfg->name;
370 iface_hint->br_type = br_cfg->datapath_type;
371 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
373 shash_add(&iface_hints, if_cfg->name, iface_hint);
379 ofproto_init(&iface_hints);
381 shash_destroy_free_data(&iface_hints);
386 if_change_cb(void *aux OVS_UNUSED)
388 ifaces_changed = true;
391 /* Public functions. */
393 /* Initializes the bridge module, configuring it to obtain its configuration
394 * from an OVSDB server accessed over 'remote', which should be a string in a
395 * form acceptable to ovsdb_idl_create(). */
397 bridge_init(const char *remote)
399 /* Create connection to database. */
400 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
401 idl_seqno = ovsdb_idl_get_seqno(idl);
402 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
403 ovsdb_idl_verify_write_only(idl);
405 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
406 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
407 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
408 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
409 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
410 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
411 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
412 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
413 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
415 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
416 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
417 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
418 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
419 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
420 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
421 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
423 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
426 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
427 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
428 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
429 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
430 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
446 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
447 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
448 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
449 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
450 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
451 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
453 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
454 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
455 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
456 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
458 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
460 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
462 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
463 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
465 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
466 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
467 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
468 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
470 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
471 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
472 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
473 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
474 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
476 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
478 /* Register unixctl commands. */
479 unixctl_command_register("qos/show", "interface", 1, 1,
480 qos_unixctl_show, NULL);
481 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
482 bridge_unixctl_dump_flows, NULL);
483 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
484 bridge_unixctl_reconnect, NULL);
493 ifnotifier = if_notifier_create(if_change_cb, NULL);
499 struct bridge *br, *next_br;
501 if_notifier_destroy(ifnotifier);
502 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
505 ovsdb_idl_destroy(idl);
508 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
509 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
510 * responsible for freeing '*managersp' (with free()).
512 * You may be asking yourself "why does ovs-vswitchd care?", because
513 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
514 * should not be and in fact is not directly involved in that. But
515 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
516 * it has to tell in-band control where the managers are to enable that.
517 * (Thus, only managers connected in-band and with non-loopback addresses
521 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
522 struct sockaddr_in **managersp, size_t *n_managersp)
524 struct sockaddr_in *managers = NULL;
525 size_t n_managers = 0;
529 /* Collect all of the potential targets from the "targets" columns of the
530 * rows pointed to by "manager_options", excluding any that are
533 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
534 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
536 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
537 sset_find_and_delete(&targets, m->target);
539 sset_add(&targets, m->target);
543 /* Now extract the targets' IP addresses. */
544 if (!sset_is_empty(&targets)) {
547 managers = xmalloc(sset_count(&targets) * sizeof *managers);
548 SSET_FOR_EACH (target, &targets) {
550 struct sockaddr_storage ss;
551 struct sockaddr_in in;
554 /* Ignore loopback. */
555 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
557 && sa.ss.ss_family == AF_INET
558 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
559 managers[n_managers++] = sa.in;
563 sset_destroy(&targets);
565 *managersp = managers;
566 *n_managersp = n_managers;
570 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
572 unsigned long int *splinter_vlans;
573 struct sockaddr_in *managers;
574 struct bridge *br, *next;
575 int sflow_bridge_number;
578 COVERAGE_INC(bridge_reconfigure);
580 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
581 OFPROTO_FLOW_LIMIT_DEFAULT));
582 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
583 OFPROTO_MAX_IDLE_DEFAULT));
584 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
586 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
589 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
590 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
592 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
593 * to 'ovs_cfg', with only very minimal configuration otherwise.
595 * This is mostly an update to bridge data structures. Nothing is pushed
596 * down to ofproto or lower layers. */
597 add_del_bridges(ovs_cfg);
598 splinter_vlans = collect_splinter_vlans(ovs_cfg);
599 HMAP_FOR_EACH (br, node, &all_bridges) {
600 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
601 bridge_del_ports(br, &br->wanted_ports);
603 free(splinter_vlans);
605 /* Start pushing configuration changes down to the ofproto layer:
607 * - Delete ofprotos that are no longer configured.
609 * - Delete ports that are no longer configured.
611 * - Reconfigure existing ports to their desired configurations, or
612 * delete them if not possible.
614 * We have to do all the deletions before we can do any additions, because
615 * the ports to be added might require resources that will be freed up by
616 * deletions (they might especially overlap in name). */
617 bridge_delete_ofprotos();
618 HMAP_FOR_EACH (br, node, &all_bridges) {
620 bridge_delete_or_reconfigure_ports(br);
624 /* Finish pushing configuration changes to the ofproto layer:
626 * - Create ofprotos that are missing.
628 * - Add ports that are missing. */
629 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
633 error = ofproto_create(br->name, br->type, &br->ofproto);
635 VLOG_ERR("failed to create bridge %s: %s", br->name,
636 ovs_strerror(error));
637 shash_destroy(&br->wanted_ports);
640 /* Trigger storing datapath version. */
641 seq_change(connectivity_seq_get());
645 HMAP_FOR_EACH (br, node, &all_bridges) {
646 bridge_add_ports(br, &br->wanted_ports);
647 shash_destroy(&br->wanted_ports);
650 reconfigure_system_stats(ovs_cfg);
652 /* Complete the configuration. */
653 sflow_bridge_number = 0;
654 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
655 HMAP_FOR_EACH (br, node, &all_bridges) {
658 /* We need the datapath ID early to allow LACP ports to use it as the
659 * default system ID. */
660 bridge_configure_datapath_id(br);
662 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
665 port_configure(port);
667 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
668 iface_set_ofport(iface->cfg, iface->ofp_port);
669 /* Clear eventual previous errors */
670 ovsrec_interface_set_error(iface->cfg, NULL);
671 iface_configure_cfm(iface);
672 iface_configure_qos(iface, port->cfg->qos);
673 iface_set_mac(br, port, iface);
674 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
676 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
680 bridge_configure_mirrors(br);
681 bridge_configure_forward_bpdu(br);
682 bridge_configure_mac_table(br);
683 bridge_configure_mcast_snooping(br);
684 bridge_configure_remotes(br, managers, n_managers);
685 bridge_configure_netflow(br);
686 bridge_configure_sflow(br, &sflow_bridge_number);
687 bridge_configure_ipfix(br);
688 bridge_configure_spanning_tree(br);
689 bridge_configure_tables(br);
690 bridge_configure_dp_desc(br);
691 bridge_configure_aa(br);
695 /* The ofproto-dpif provider does some final reconfiguration in its
696 * ->type_run() function. We have to call it before notifying the database
697 * client that reconfiguration is complete, otherwise there is a very
698 * narrow race window in which e.g. ofproto/trace will not recognize the
699 * new configuration (sometimes this causes unit test failures). */
703 /* Delete ofprotos which aren't configured or have the wrong type. Create
704 * ofprotos which don't exist but need to. */
706 bridge_delete_ofprotos(void)
713 /* Delete ofprotos with no bridge or with the wrong type. */
716 ofproto_enumerate_types(&types);
717 SSET_FOR_EACH (type, &types) {
720 ofproto_enumerate_names(type, &names);
721 SSET_FOR_EACH (name, &names) {
722 br = bridge_lookup(name);
723 if (!br || strcmp(type, br->type)) {
724 ofproto_delete(name, type);
728 sset_destroy(&names);
729 sset_destroy(&types);
733 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
735 if (*n >= *allocated) {
736 ports = x2nrealloc(ports, allocated, sizeof *ports);
738 ports[(*n)++] = port;
743 bridge_delete_or_reconfigure_ports(struct bridge *br)
745 struct ofproto_port ofproto_port;
746 struct ofproto_port_dump dump;
748 struct sset ofproto_ports;
749 struct port *port, *port_next;
751 /* List of "ofp_port"s to delete. We make a list instead of deleting them
752 * right away because ofproto implementations aren't necessarily able to
753 * iterate through a changing list of ports in an entirely robust way. */
760 sset_init(&ofproto_ports);
762 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
763 * that are not configured in the database. (This commonly happens when
764 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
766 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
767 * that have the wrong OpenFlow port number (and arrange to add them back
768 * with the correct OpenFlow port number). */
769 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
770 ofp_port_t requested_ofp_port;
773 sset_add(&ofproto_ports, ofproto_port.name);
775 iface = iface_lookup(br, ofproto_port.name);
777 /* No such iface is configured, so we should delete this
780 * As a corner case exception, keep the port if it's a bond fake
782 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
783 && !strcmp(ofproto_port.type, "internal")) {
789 if (strcmp(ofproto_port.type, iface->type)
790 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
791 /* The interface is the wrong type or can't be configured.
796 /* If the requested OpenFlow port for 'iface' changed, and it's not
797 * already the correct port, then we might want to temporarily delete
798 * this interface, so we can add it back again with the new OpenFlow
800 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
801 if (iface->ofp_port != OFPP_LOCAL &&
802 requested_ofp_port != OFPP_NONE &&
803 requested_ofp_port != iface->ofp_port) {
804 ofp_port_t victim_request;
805 struct iface *victim;
807 /* Check for an existing OpenFlow port currently occupying
808 * 'iface''s requested port number. If there isn't one, then
809 * delete this port. Otherwise we need to consider further. */
810 victim = iface_from_ofp_port(br, requested_ofp_port);
815 /* 'victim' is a port currently using 'iface''s requested port
816 * number. Unless 'victim' specifically requested that port
817 * number, too, then we can delete both 'iface' and 'victim'
818 * temporarily. (We'll add both of them back again later with new
819 * OpenFlow port numbers.)
821 * If 'victim' did request port number 'requested_ofp_port', just
822 * like 'iface', then that's a configuration inconsistency that we
823 * can't resolve. We might as well let it keep its current port
825 victim_request = iface_get_requested_ofp_port(victim->cfg);
826 if (victim_request != requested_ofp_port) {
827 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
828 iface_destroy(victim);
837 iface_destroy(iface);
838 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
840 for (i = 0; i < n; i++) {
841 ofproto_port_del(br->ofproto, del[i]);
845 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
846 * that we didn't see when we iterated through the datapath, i.e. ports
847 * that disappeared underneath use. This is an unusual situation, but it
848 * can happen in some cases:
850 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
851 * idea but could happen).
853 * - The port represented a device that disappeared, e.g. a tuntap
854 * device destroyed via "tunctl -d", a physical Ethernet device
855 * whose module was just unloaded via "rmmod", or a virtual NIC for a
856 * VM whose VM was just terminated. */
857 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
858 struct iface *iface, *iface_next;
860 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
861 if (!sset_contains(&ofproto_ports, iface->name)) {
862 iface_destroy__(iface);
866 if (list_is_empty(&port->ifaces)) {
870 sset_destroy(&ofproto_ports);
874 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
875 bool with_requested_port)
877 struct shash_node *port_node;
879 SHASH_FOR_EACH (port_node, wanted_ports) {
880 const struct ovsrec_port *port_cfg = port_node->data;
883 for (i = 0; i < port_cfg->n_interfaces; i++) {
884 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
885 ofp_port_t requested_ofp_port;
887 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
888 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
889 struct iface *iface = iface_lookup(br, iface_cfg->name);
892 iface_create(br, iface_cfg, port_cfg);
900 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
902 /* First add interfaces that request a particular port number. */
903 bridge_add_ports__(br, wanted_ports, true);
905 /* Then add interfaces that want automatic port number assignment.
906 * We add these afterward to avoid accidentally taking a specifically
907 * requested port number. */
908 bridge_add_ports__(br, wanted_ports, false);
912 port_configure(struct port *port)
914 const struct ovsrec_port *cfg = port->cfg;
915 struct bond_settings bond_settings;
916 struct lacp_settings lacp_settings;
917 struct ofproto_bundle_settings s;
920 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
921 configure_splinter_port(port);
930 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
931 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
932 s.slaves[s.n_slaves++] = iface->ofp_port;
937 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
941 /* Get VLAN trunks. */
944 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
948 if (cfg->vlan_mode) {
949 if (!strcmp(cfg->vlan_mode, "access")) {
950 s.vlan_mode = PORT_VLAN_ACCESS;
951 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
952 s.vlan_mode = PORT_VLAN_TRUNK;
953 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
954 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
955 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
956 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
958 /* This "can't happen" because ovsdb-server should prevent it. */
959 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
960 "back to trunk mode", port->name, cfg->vlan_mode);
961 s.vlan_mode = PORT_VLAN_TRUNK;
965 s.vlan_mode = PORT_VLAN_ACCESS;
967 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
971 s.vlan_mode = PORT_VLAN_TRUNK;
974 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
977 /* Get LACP settings. */
978 s.lacp = port_configure_lacp(port, &lacp_settings);
982 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
983 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
984 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
987 s.lacp_slaves = NULL;
990 /* Get bond settings. */
991 if (s.n_slaves > 1) {
992 s.bond = &bond_settings;
993 port_configure_bond(port, &bond_settings);
996 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
997 netdev_set_miimon_interval(iface->netdev, 0);
1002 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1007 free(s.lacp_slaves);
1010 /* Pick local port hardware address and datapath ID for 'br'. */
1012 bridge_configure_datapath_id(struct bridge *br)
1016 struct iface *local_iface;
1017 struct iface *hw_addr_iface;
1020 bridge_pick_local_hw_addr(br, &ea, &hw_addr_iface);
1021 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1023 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1025 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1026 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1027 "Ethernet address: %s",
1028 br->name, ovs_strerror(error));
1033 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1034 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1035 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1036 ofproto_set_datapath_id(br->ofproto, dpid);
1039 dpid_string = xasprintf("%016"PRIx64, dpid);
1040 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1044 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1047 bridge_get_allowed_versions(struct bridge *br)
1049 if (!br->cfg->n_protocols) {
1053 return ofputil_versions_from_strings(br->cfg->protocols,
1054 br->cfg->n_protocols);
1057 /* Set NetFlow configuration on 'br'. */
1059 bridge_configure_netflow(struct bridge *br)
1061 struct ovsrec_netflow *cfg = br->cfg->netflow;
1062 struct netflow_options opts;
1065 ofproto_set_netflow(br->ofproto, NULL);
1069 memset(&opts, 0, sizeof opts);
1071 /* Get default NetFlow configuration from datapath.
1072 * Apply overrides from 'cfg'. */
1073 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1074 if (cfg->engine_type) {
1075 opts.engine_type = *cfg->engine_type;
1077 if (cfg->engine_id) {
1078 opts.engine_id = *cfg->engine_id;
1081 /* Configure active timeout interval. */
1082 opts.active_timeout = cfg->active_timeout;
1083 if (!opts.active_timeout) {
1084 opts.active_timeout = -1;
1085 } else if (opts.active_timeout < 0) {
1086 VLOG_WARN("bridge %s: active timeout interval set to negative "
1087 "value, using default instead (%d seconds)", br->name,
1088 NF_ACTIVE_TIMEOUT_DEFAULT);
1089 opts.active_timeout = -1;
1092 /* Add engine ID to interface number to disambiguate bridgs? */
1093 opts.add_id_to_iface = cfg->add_id_to_interface;
1094 if (opts.add_id_to_iface) {
1095 if (opts.engine_id > 0x7f) {
1096 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1097 "another vswitch, choose an engine id less than 128",
1100 if (hmap_count(&br->ports) > 508) {
1101 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1102 "another port when more than 508 ports are used",
1108 sset_init(&opts.collectors);
1109 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1112 if (ofproto_set_netflow(br->ofproto, &opts)) {
1113 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1115 sset_destroy(&opts.collectors);
1118 /* Set sFlow configuration on 'br'. */
1120 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1122 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1123 struct ovsrec_controller **controllers;
1124 struct ofproto_sflow_options oso;
1125 size_t n_controllers;
1129 ofproto_set_sflow(br->ofproto, NULL);
1133 memset(&oso, 0, sizeof oso);
1135 sset_init(&oso.targets);
1136 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1138 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1139 if (cfg->sampling) {
1140 oso.sampling_rate = *cfg->sampling;
1143 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1145 oso.polling_interval = *cfg->polling;
1148 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1150 oso.header_len = *cfg->header;
1153 oso.sub_id = (*sflow_bridge_number)++;
1154 oso.agent_device = cfg->agent;
1156 oso.control_ip = NULL;
1157 n_controllers = bridge_get_controllers(br, &controllers);
1158 for (i = 0; i < n_controllers; i++) {
1159 if (controllers[i]->local_ip) {
1160 oso.control_ip = controllers[i]->local_ip;
1164 ofproto_set_sflow(br->ofproto, &oso);
1166 sset_destroy(&oso.targets);
1169 /* Returns whether a IPFIX row is valid. */
1171 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1173 return ipfix && ipfix->n_targets > 0;
1176 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1178 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1179 const struct bridge *br)
1181 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1184 /* Set IPFIX configuration on 'br'. */
1186 bridge_configure_ipfix(struct bridge *br)
1188 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1189 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1190 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1191 struct ofproto_ipfix_bridge_exporter_options be_opts;
1192 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1193 size_t n_fe_opts = 0;
1195 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1196 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1201 if (!valid_be_cfg && n_fe_opts == 0) {
1202 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1207 memset(&be_opts, 0, sizeof be_opts);
1209 sset_init(&be_opts.targets);
1210 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1212 if (be_cfg->sampling) {
1213 be_opts.sampling_rate = *be_cfg->sampling;
1215 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1217 if (be_cfg->obs_domain_id) {
1218 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1220 if (be_cfg->obs_point_id) {
1221 be_opts.obs_point_id = *be_cfg->obs_point_id;
1223 if (be_cfg->cache_active_timeout) {
1224 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1226 if (be_cfg->cache_max_flows) {
1227 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1230 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1231 "enable-tunnel-sampling", true);
1233 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1234 "enable-input-sampling", false);
1236 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1237 "enable-output-sampling", false);
1240 if (n_fe_opts > 0) {
1241 struct ofproto_ipfix_flow_exporter_options *opts;
1242 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1244 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1245 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1246 opts->collector_set_id = fe_cfg->id;
1247 sset_init(&opts->targets);
1248 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1249 fe_cfg->ipfix->n_targets);
1250 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1251 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1252 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1253 ? *fe_cfg->ipfix->cache_max_flows : 0;
1259 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1263 sset_destroy(&be_opts.targets);
1266 if (n_fe_opts > 0) {
1267 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1269 for (i = 0; i < n_fe_opts; i++) {
1270 sset_destroy(&opts->targets);
1278 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1279 struct ofproto_port_stp_settings *port_s,
1280 int *port_num_counter, unsigned long *port_num_bitmap)
1282 const char *config_str;
1283 struct iface *iface;
1285 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1286 port_s->enable = false;
1289 port_s->enable = true;
1292 /* STP over bonds is not supported. */
1293 if (!list_is_singleton(&port->ifaces)) {
1294 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1296 port_s->enable = false;
1300 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1302 /* Internal ports shouldn't participate in spanning tree, so
1304 if (!strcmp(iface->type, "internal")) {
1305 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1306 port_s->enable = false;
1310 /* STP on mirror output ports is not supported. */
1311 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1312 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1313 port_s->enable = false;
1317 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1319 unsigned long int port_num = strtoul(config_str, NULL, 0);
1320 int port_idx = port_num - 1;
1322 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1323 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1324 port_s->enable = false;
1328 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1329 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1330 port->name, port_num);
1331 port_s->enable = false;
1334 bitmap_set1(port_num_bitmap, port_idx);
1335 port_s->port_num = port_idx;
1337 if (*port_num_counter >= STP_MAX_PORTS) {
1338 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1339 port_s->enable = false;
1343 port_s->port_num = (*port_num_counter)++;
1346 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1348 port_s->path_cost = strtoul(config_str, NULL, 10);
1350 enum netdev_features current;
1353 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1354 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1355 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1358 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1360 port_s->priority = strtoul(config_str, NULL, 0);
1362 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1367 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1368 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1370 const char *config_str;
1371 struct iface *iface;
1373 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1374 port_s->enable = false;
1377 port_s->enable = true;
1380 /* RSTP over bonds is not supported. */
1381 if (!list_is_singleton(&port->ifaces)) {
1382 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1384 port_s->enable = false;
1388 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1390 /* Internal ports shouldn't participate in spanning tree, so
1392 if (!strcmp(iface->type, "internal")) {
1393 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1394 port_s->enable = false;
1398 /* RSTP on mirror output ports is not supported. */
1399 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1400 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1401 port_s->enable = false;
1405 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1407 unsigned long int port_num = strtoul(config_str, NULL, 0);
1408 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1409 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1410 port_s->enable = false;
1413 port_s->port_num = port_num;
1415 if (*port_num_counter >= RSTP_MAX_PORTS) {
1416 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1417 port_s->enable = false;
1420 /* If rstp-port-num is not specified, use 0.
1421 * rstp_port_set_port_number() will look for the first free one. */
1422 port_s->port_num = 0;
1425 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1427 port_s->path_cost = strtoul(config_str, NULL, 10);
1429 enum netdev_features current;
1432 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1433 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1434 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1437 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1439 port_s->priority = strtoul(config_str, NULL, 0);
1441 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1444 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1446 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1448 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1451 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1452 "rstp-admin-port-state", true);
1454 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1455 "rstp-port-admin-edge", false);
1456 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1457 "rstp-port-auto-edge", true);
1458 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1459 "rstp-port-mcheck", false);
1462 /* Set spanning tree configuration on 'br'. */
1464 bridge_configure_stp(struct bridge *br, bool enable_stp)
1467 ofproto_set_stp(br->ofproto, NULL);
1469 struct ofproto_stp_settings br_s;
1470 const char *config_str;
1472 int port_num_counter;
1473 unsigned long *port_num_bitmap;
1475 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1479 if (eth_addr_from_string(config_str, &ea)) {
1480 br_s.system_id = eth_addr_to_uint64(ea);
1482 br_s.system_id = eth_addr_to_uint64(br->ea);
1483 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1484 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1487 br_s.system_id = eth_addr_to_uint64(br->ea);
1490 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1492 br_s.priority = strtoul(config_str, NULL, 0);
1494 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1497 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1499 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1501 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1504 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1506 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1508 br_s.max_age = STP_DEFAULT_MAX_AGE;
1511 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1513 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1515 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1518 /* Configure STP on the bridge. */
1519 if (ofproto_set_stp(br->ofproto, &br_s)) {
1520 VLOG_ERR("bridge %s: could not enable STP", br->name);
1524 /* Users must either set the port number with the "stp-port-num"
1525 * configuration on all ports or none. If manual configuration
1526 * is not done, then we allocate them sequentially. */
1527 port_num_counter = 0;
1528 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1529 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1530 struct ofproto_port_stp_settings port_s;
1531 struct iface *iface;
1533 port_configure_stp(br->ofproto, port, &port_s,
1534 &port_num_counter, port_num_bitmap);
1536 /* As bonds are not supported, just apply configuration to
1537 * all interfaces. */
1538 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1539 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1541 VLOG_ERR("port %s: could not enable STP", port->name);
1547 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1548 && port_num_counter) {
1549 VLOG_ERR("bridge %s: must manually configure all STP port "
1550 "IDs or none, disabling", br->name);
1551 ofproto_set_stp(br->ofproto, NULL);
1553 bitmap_free(port_num_bitmap);
1558 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1561 ofproto_set_rstp(br->ofproto, NULL);
1563 struct ofproto_rstp_settings br_s;
1564 const char *config_str;
1566 int port_num_counter;
1568 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1572 if (eth_addr_from_string(config_str, &ea)) {
1573 br_s.address = eth_addr_to_uint64(ea);
1576 br_s.address = eth_addr_to_uint64(br->ea);
1577 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1578 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1582 br_s.address = eth_addr_to_uint64(br->ea);
1585 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1587 br_s.priority = strtoul(config_str, NULL, 0);
1589 br_s.priority = RSTP_DEFAULT_PRIORITY;
1592 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1594 br_s.ageing_time = strtoul(config_str, NULL, 0);
1596 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1599 config_str = smap_get(&br->cfg->other_config,
1600 "rstp-force-protocol-version");
1602 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1604 br_s.force_protocol_version = FPV_DEFAULT;
1607 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1609 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1611 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1614 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1616 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1618 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1621 config_str = smap_get(&br->cfg->other_config,
1622 "rstp-transmit-hold-count");
1624 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1626 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1629 /* Configure RSTP on the bridge. */
1630 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1631 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1635 port_num_counter = 0;
1636 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1637 struct ofproto_port_rstp_settings port_s;
1638 struct iface *iface;
1640 port_configure_rstp(br->ofproto, port, &port_s,
1643 /* As bonds are not supported, just apply configuration to
1644 * all interfaces. */
1645 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1646 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1648 VLOG_ERR("port %s: could not enable RSTP", port->name);
1657 bridge_configure_spanning_tree(struct bridge *br)
1659 bool enable_rstp = br->cfg->rstp_enable;
1660 bool enable_stp = br->cfg->stp_enable;
1662 if (enable_rstp && enable_stp) {
1663 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1664 "configured; enabling RSTP", br->name);
1668 bridge_configure_stp(br, enable_stp);
1669 bridge_configure_rstp(br, enable_rstp);
1673 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1675 const struct port *port = port_lookup(br, name);
1676 return port && port_is_bond_fake_iface(port);
1680 port_is_bond_fake_iface(const struct port *port)
1682 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1686 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1688 struct bridge *br, *next;
1689 struct shash new_br;
1692 /* Collect new bridges' names and types. */
1693 shash_init(&new_br);
1694 for (i = 0; i < cfg->n_bridges; i++) {
1695 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1696 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1698 if (strchr(br_cfg->name, '/')) {
1699 /* Prevent remote ovsdb-server users from accessing arbitrary
1700 * directories, e.g. consider a bridge named "../../../etc/". */
1701 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1703 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1704 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1708 /* Get rid of deleted bridges or those whose types have changed.
1709 * Update 'cfg' of bridges that still exist. */
1710 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1711 br->cfg = shash_find_data(&new_br, br->name);
1712 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1713 br->cfg->datapath_type))) {
1718 /* Add new bridges. */
1719 for (i = 0; i < cfg->n_bridges; i++) {
1720 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1721 struct bridge *br = bridge_lookup(br_cfg->name);
1723 bridge_create(br_cfg);
1727 shash_destroy(&new_br);
1730 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1731 * Returns 0 if successful, otherwise a positive errno value. */
1733 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1734 struct netdev *netdev, char **errp)
1736 return netdev_set_config(netdev, &iface_cfg->options, errp);
1739 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1740 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1742 * If successful, returns 0 and stores the network device in '*netdevp'. On
1743 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1745 iface_do_create(const struct bridge *br,
1746 const struct ovsrec_interface *iface_cfg,
1747 const struct ovsrec_port *port_cfg,
1748 ofp_port_t *ofp_portp, struct netdev **netdevp,
1751 struct netdev *netdev = NULL;
1754 if (netdev_is_reserved_name(iface_cfg->name)) {
1755 VLOG_WARN("could not create interface %s, name is reserved",
1761 error = netdev_open(iface_cfg->name,
1762 iface_get_type(iface_cfg, br->cfg), &netdev);
1764 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1765 iface_cfg->name, ovs_strerror(error));
1769 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1774 *ofp_portp = iface_pick_ofport(iface_cfg);
1775 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1780 VLOG_INFO("bridge %s: added interface %s on port %d",
1781 br->name, iface_cfg->name, *ofp_portp);
1783 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1784 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1792 netdev_close(netdev);
1796 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1797 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1798 * automatically allocated for the iface. Takes ownership of and
1799 * deallocates 'if_cfg'.
1801 * Return true if an iface is successfully created, false otherwise. */
1803 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1804 const struct ovsrec_port *port_cfg)
1806 struct netdev *netdev;
1807 struct iface *iface;
1808 ofp_port_t ofp_port;
1813 /* Do the bits that can fail up front. */
1814 ovs_assert(!iface_lookup(br, iface_cfg->name));
1815 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1817 iface_clear_db_record(iface_cfg, errp);
1822 /* Get or create the port structure. */
1823 port = port_lookup(br, port_cfg->name);
1825 port = port_create(br, port_cfg);
1828 /* Create the iface structure. */
1829 iface = xzalloc(sizeof *iface);
1830 list_push_back(&port->ifaces, &iface->port_elem);
1831 hmap_insert(&br->iface_by_name, &iface->name_node,
1832 hash_string(iface_cfg->name, 0));
1834 iface->name = xstrdup(iface_cfg->name);
1835 iface->ofp_port = ofp_port;
1836 iface->netdev = netdev;
1837 iface->type = iface_get_type(iface_cfg, br->cfg);
1838 iface->cfg = iface_cfg;
1839 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1840 hash_ofp_port(ofp_port));
1842 /* Populate initial status in database. */
1843 iface_refresh_stats(iface);
1844 iface_refresh_netdev_status(iface);
1846 /* Add bond fake iface if necessary. */
1847 if (port_is_bond_fake_iface(port)) {
1848 struct ofproto_port ofproto_port;
1850 if (ofproto_port_query_by_name(br->ofproto, port->name,
1852 struct netdev *netdev;
1855 error = netdev_open(port->name, "internal", &netdev);
1857 ofp_port_t fake_ofp_port = OFPP_NONE;
1858 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1859 netdev_close(netdev);
1861 VLOG_WARN("could not open network device %s (%s)",
1862 port->name, ovs_strerror(error));
1865 /* Already exists, nothing to do. */
1866 ofproto_port_destroy(&ofproto_port);
1873 /* Set forward BPDU option. */
1875 bridge_configure_forward_bpdu(struct bridge *br)
1877 ofproto_set_forward_bpdu(br->ofproto,
1878 smap_get_bool(&br->cfg->other_config,
1883 /* Set MAC learning table configuration for 'br'. */
1885 bridge_configure_mac_table(struct bridge *br)
1887 const char *idle_time_str;
1890 const char *mac_table_size_str;
1893 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1894 idle_time = (idle_time_str && atoi(idle_time_str)
1895 ? atoi(idle_time_str)
1896 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1898 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1899 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1900 ? atoi(mac_table_size_str)
1903 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1906 /* Set multicast snooping table configuration for 'br'. */
1908 bridge_configure_mcast_snooping(struct bridge *br)
1910 if (!br->cfg->mcast_snooping_enable) {
1911 ofproto_set_mcast_snooping(br->ofproto, NULL);
1914 struct ofproto_mcast_snooping_settings br_s;
1915 const char *idle_time_str;
1916 const char *max_entries_str;
1918 idle_time_str = smap_get(&br->cfg->other_config,
1919 "mcast-snooping-aging-time");
1920 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1921 ? atoi(idle_time_str)
1922 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1924 max_entries_str = smap_get(&br->cfg->other_config,
1925 "mcast-snooping-table-size");
1926 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1927 ? atoi(max_entries_str)
1928 : MCAST_DEFAULT_MAX_ENTRIES);
1930 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1931 "mcast-snooping-disable-flood-unregistered",
1934 /* Configure multicast snooping on the bridge */
1935 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1936 VLOG_ERR("bridge %s: could not enable multicast snooping",
1941 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1942 struct ofproto_mcast_snooping_port_settings port_s;
1943 port_s.flood = smap_get_bool(&port->cfg->other_config,
1944 "mcast-snooping-flood", false);
1945 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1946 "mcast-snooping-flood-reports", false);
1947 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1948 VLOG_ERR("port %s: could not configure mcast snooping",
1956 find_local_hw_addr(const struct bridge *br, struct eth_addr *ea,
1957 const struct port *fake_br, struct iface **hw_addr_iface)
1959 struct hmapx mirror_output_ports;
1961 bool found_addr = false;
1965 /* Mirror output ports don't participate in picking the local hardware
1966 * address. ofproto can't help us find out whether a given port is a
1967 * mirror output because we haven't configured mirrors yet, so we need to
1968 * accumulate them ourselves. */
1969 hmapx_init(&mirror_output_ports);
1970 for (i = 0; i < br->cfg->n_mirrors; i++) {
1971 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1972 if (m->output_port) {
1973 hmapx_add(&mirror_output_ports, m->output_port);
1977 /* Otherwise choose the minimum non-local MAC address among all of the
1979 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1980 struct eth_addr iface_ea;
1981 struct iface *candidate;
1982 struct iface *iface;
1984 /* Mirror output ports don't participate. */
1985 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1989 /* Choose the MAC address to represent the port. */
1991 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac,
1993 /* Find the interface with this Ethernet address (if any) so that
1994 * we can provide the correct devname to the caller. */
1995 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1996 struct eth_addr candidate_ea;
1997 if (!netdev_get_etheraddr(candidate->netdev, &candidate_ea)
1998 && eth_addr_equals(iface_ea, candidate_ea)) {
2003 /* Choose the interface whose MAC address will represent the port.
2004 * The Linux kernel bonding code always chooses the MAC address of
2005 * the first slave added to a bond, and the Fedora networking
2006 * scripts always add slaves to a bond in alphabetical order, so
2007 * for compatibility we choose the interface with the name that is
2008 * first in alphabetical order. */
2009 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2010 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2015 /* The local port doesn't count (since we're trying to choose its
2016 * MAC address anyway). */
2017 if (iface->ofp_port == OFPP_LOCAL) {
2021 /* For fake bridges we only choose from ports with the same tag */
2022 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2023 if (!port->cfg->tag) {
2026 if (*port->cfg->tag != *fake_br->cfg->tag) {
2032 error = netdev_get_etheraddr(iface->netdev, &iface_ea);
2038 /* Compare against our current choice. */
2039 if (!eth_addr_is_multicast(iface_ea) &&
2040 !eth_addr_is_local(iface_ea) &&
2041 !eth_addr_is_reserved(iface_ea) &&
2042 !eth_addr_is_zero(iface_ea) &&
2043 (!found_addr || eth_addr_compare_3way(iface_ea, *ea) < 0))
2046 *hw_addr_iface = iface;
2052 *ea = br->default_ea;
2053 *hw_addr_iface = NULL;
2056 hmapx_destroy(&mirror_output_ports);
2060 bridge_pick_local_hw_addr(struct bridge *br, struct eth_addr *ea,
2061 struct iface **hw_addr_iface)
2064 *hw_addr_iface = NULL;
2066 /* Did the user request a particular MAC? */
2067 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2068 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2069 if (eth_addr_is_multicast(*ea)) {
2070 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2071 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(*ea));
2072 } else if (eth_addr_is_zero(*ea)) {
2073 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2079 /* Find a local hw address */
2080 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2083 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2084 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2085 * an interface on 'br', then that interface must be passed in as
2086 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2087 * 'hw_addr_iface' must be passed in as a null pointer. */
2089 bridge_pick_datapath_id(struct bridge *br,
2090 const struct eth_addr bridge_ea,
2091 struct iface *hw_addr_iface)
2094 * The procedure for choosing a bridge MAC address will, in the most
2095 * ordinary case, also choose a unique MAC that we can use as a datapath
2096 * ID. In some special cases, though, multiple bridges will end up with
2097 * the same MAC address. This is OK for the bridges, but it will confuse
2098 * the OpenFlow controller, because each datapath needs a unique datapath
2101 * Datapath IDs must be unique. It is also very desirable that they be
2102 * stable from one run to the next, so that policy set on a datapath
2105 const char *datapath_id;
2108 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2109 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2113 if (!hw_addr_iface) {
2115 * A purely internal bridge, that is, one that has no non-virtual
2116 * network devices on it at all, is difficult because it has no
2117 * natural unique identifier at all.
2119 * When the host is a XenServer, we handle this case by hashing the
2120 * host's UUID with the name of the bridge. Names of bridges are
2121 * persistent across XenServer reboots, although they can be reused if
2122 * an internal network is destroyed and then a new one is later
2123 * created, so this is fairly effective.
2125 * When the host is not a XenServer, we punt by using a random MAC
2126 * address on each run.
2128 const char *host_uuid = xenserver_get_host_uuid();
2130 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2131 dpid = dpid_from_hash(combined, strlen(combined));
2137 return eth_addr_to_uint64(bridge_ea);
2141 dpid_from_hash(const void *data, size_t n)
2144 uint8_t bytes[SHA1_DIGEST_SIZE];
2148 sha1_bytes(data, n, hash.bytes);
2149 eth_addr_mark_random(&hash.ea);
2150 return eth_addr_to_uint64(hash.ea);
2154 iface_refresh_netdev_status(struct iface *iface)
2158 enum netdev_features current;
2159 enum netdev_flags flags;
2160 const char *link_state;
2161 struct eth_addr mac;
2162 int64_t bps, mtu_64, ifindex64, link_resets;
2165 if (iface_is_synthetic(iface)) {
2169 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2170 && !status_txn_try_again) {
2174 iface->change_seq = netdev_get_change_seq(iface->netdev);
2178 if (!netdev_get_status(iface->netdev, &smap)) {
2179 ovsrec_interface_set_status(iface->cfg, &smap);
2181 ovsrec_interface_set_status(iface->cfg, NULL);
2184 smap_destroy(&smap);
2186 error = netdev_get_flags(iface->netdev, &flags);
2188 const char *state = flags & NETDEV_UP ? "up" : "down";
2190 ovsrec_interface_set_admin_state(iface->cfg, state);
2192 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2195 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2196 ovsrec_interface_set_link_state(iface->cfg, link_state);
2198 link_resets = netdev_get_carrier_resets(iface->netdev);
2199 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2201 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2202 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2204 ovsrec_interface_set_duplex(iface->cfg,
2205 netdev_features_is_full_duplex(current)
2207 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2209 ovsrec_interface_set_duplex(iface->cfg, NULL);
2210 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2213 error = netdev_get_mtu(iface->netdev, &mtu);
2216 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2218 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2221 error = netdev_get_etheraddr(iface->netdev, &mac);
2223 char mac_string[32];
2225 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2226 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2228 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2231 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2232 * if there is no valid ifindex number. */
2233 ifindex64 = netdev_get_ifindex(iface->netdev);
2234 if (ifindex64 < 0) {
2237 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2241 iface_refresh_ofproto_status(struct iface *iface)
2245 if (iface_is_synthetic(iface)) {
2249 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2253 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2255 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2258 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2260 || status_txn_try_again) {
2261 iface_refresh_cfm_stats(iface);
2264 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2266 || status_txn_try_again) {
2270 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2271 iface->ofp_port, &smap);
2272 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2273 smap_destroy(&smap);
2277 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2280 iface_refresh_cfm_stats(struct iface *iface)
2282 const struct ovsrec_interface *cfg = iface->cfg;
2283 struct cfm_status status;
2286 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2287 iface->ofp_port, &status);
2289 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2290 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2291 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2292 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2293 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2294 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2296 const char *reasons[CFM_FAULT_N_REASONS];
2297 int64_t cfm_health = status.health;
2298 int64_t cfm_flap_count = status.flap_count;
2299 bool faulted = status.faults != 0;
2302 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2305 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2306 int reason = 1 << i;
2307 if (status.faults & reason) {
2308 reasons[j++] = cfm_fault_reason_to_str(reason);
2311 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2313 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2315 if (status.remote_opstate >= 0) {
2316 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2317 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2319 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2322 ovsrec_interface_set_cfm_remote_mpids(cfg,
2323 (const int64_t *)status.rmps,
2325 if (cfm_health >= 0) {
2326 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2328 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2336 iface_refresh_stats(struct iface *iface)
2338 #define IFACE_STATS \
2339 IFACE_STAT(rx_packets, "rx_packets") \
2340 IFACE_STAT(tx_packets, "tx_packets") \
2341 IFACE_STAT(rx_bytes, "rx_bytes") \
2342 IFACE_STAT(tx_bytes, "tx_bytes") \
2343 IFACE_STAT(rx_dropped, "rx_dropped") \
2344 IFACE_STAT(tx_dropped, "tx_dropped") \
2345 IFACE_STAT(rx_errors, "rx_errors") \
2346 IFACE_STAT(tx_errors, "tx_errors") \
2347 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2348 IFACE_STAT(rx_over_errors, "rx_over_err") \
2349 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2350 IFACE_STAT(collisions, "collisions")
2352 #define IFACE_STAT(MEMBER, NAME) + 1
2353 enum { N_IFACE_STATS = IFACE_STATS };
2355 int64_t values[N_IFACE_STATS];
2356 const char *keys[N_IFACE_STATS];
2359 struct netdev_stats stats;
2361 if (iface_is_synthetic(iface)) {
2365 /* Intentionally ignore return value, since errors will set 'stats' to
2366 * all-1s, and we will deal with that correctly below. */
2367 netdev_get_stats(iface->netdev, &stats);
2369 /* Copy statistics into keys[] and values[]. */
2371 #define IFACE_STAT(MEMBER, NAME) \
2372 if (stats.MEMBER != UINT64_MAX) { \
2374 values[n] = stats.MEMBER; \
2379 ovs_assert(n <= N_IFACE_STATS);
2381 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2386 br_refresh_datapath_info(struct bridge *br)
2388 const char *version;
2390 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2391 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2394 ovsrec_bridge_set_datapath_version(br->cfg,
2395 version ? version : "<unknown>");
2399 br_refresh_stp_status(struct bridge *br)
2401 struct smap smap = SMAP_INITIALIZER(&smap);
2402 struct ofproto *ofproto = br->ofproto;
2403 struct ofproto_stp_status status;
2405 if (ofproto_get_stp_status(ofproto, &status)) {
2409 if (!status.enabled) {
2410 ovsrec_bridge_set_status(br->cfg, NULL);
2414 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2415 STP_ID_ARGS(status.bridge_id));
2416 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2417 STP_ID_ARGS(status.designated_root));
2418 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2420 ovsrec_bridge_set_status(br->cfg, &smap);
2421 smap_destroy(&smap);
2425 port_refresh_stp_status(struct port *port)
2427 struct ofproto *ofproto = port->bridge->ofproto;
2428 struct iface *iface;
2429 struct ofproto_port_stp_status status;
2432 if (port_is_synthetic(port)) {
2436 /* STP doesn't currently support bonds. */
2437 if (!list_is_singleton(&port->ifaces)) {
2438 ovsrec_port_set_status(port->cfg, NULL);
2442 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2443 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2447 if (!status.enabled) {
2448 ovsrec_port_set_status(port->cfg, NULL);
2452 /* Set Status column. */
2454 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2455 smap_add(&smap, "stp_state", stp_state_name(status.state));
2456 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2457 smap_add(&smap, "stp_role", stp_role_name(status.role));
2458 ovsrec_port_set_status(port->cfg, &smap);
2459 smap_destroy(&smap);
2463 port_refresh_stp_stats(struct port *port)
2465 struct ofproto *ofproto = port->bridge->ofproto;
2466 struct iface *iface;
2467 struct ofproto_port_stp_stats stats;
2468 const char *keys[3];
2469 int64_t int_values[3];
2471 if (port_is_synthetic(port)) {
2475 /* STP doesn't currently support bonds. */
2476 if (!list_is_singleton(&port->ifaces)) {
2480 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2481 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2485 if (!stats.enabled) {
2486 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2490 /* Set Statistics column. */
2491 keys[0] = "stp_tx_count";
2492 int_values[0] = stats.tx_count;
2493 keys[1] = "stp_rx_count";
2494 int_values[1] = stats.rx_count;
2495 keys[2] = "stp_error_count";
2496 int_values[2] = stats.error_count;
2498 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2499 ARRAY_SIZE(int_values));
2503 br_refresh_rstp_status(struct bridge *br)
2505 struct smap smap = SMAP_INITIALIZER(&smap);
2506 struct ofproto *ofproto = br->ofproto;
2507 struct ofproto_rstp_status status;
2509 if (ofproto_get_rstp_status(ofproto, &status)) {
2512 if (!status.enabled) {
2513 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2516 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2517 RSTP_ID_ARGS(status.bridge_id));
2518 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2519 status.root_path_cost);
2520 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2521 RSTP_ID_ARGS(status.root_id));
2522 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2523 RSTP_ID_ARGS(status.designated_id));
2524 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2525 status.designated_port_id);
2526 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2527 status.bridge_port_id);
2528 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2529 smap_destroy(&smap);
2533 port_refresh_rstp_status(struct port *port)
2535 struct ofproto *ofproto = port->bridge->ofproto;
2536 struct iface *iface;
2537 struct ofproto_port_rstp_status status;
2538 const char *keys[4];
2539 int64_t int_values[4];
2542 if (port_is_synthetic(port)) {
2546 /* RSTP doesn't currently support bonds. */
2547 if (!list_is_singleton(&port->ifaces)) {
2548 ovsrec_port_set_rstp_status(port->cfg, NULL);
2552 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2553 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2557 if (!status.enabled) {
2558 ovsrec_port_set_rstp_status(port->cfg, NULL);
2559 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2562 /* Set Status column. */
2565 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2567 smap_add_format(&smap, "rstp_port_role", "%s",
2568 rstp_port_role_name(status.role));
2569 smap_add_format(&smap, "rstp_port_state", "%s",
2570 rstp_state_name(status.state));
2571 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2572 RSTP_ID_ARGS(status.designated_bridge_id));
2573 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2574 status.designated_port_id);
2575 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2576 status.designated_path_cost);
2578 ovsrec_port_set_rstp_status(port->cfg, &smap);
2579 smap_destroy(&smap);
2581 /* Set Statistics column. */
2582 keys[0] = "rstp_tx_count";
2583 int_values[0] = status.tx_count;
2584 keys[1] = "rstp_rx_count";
2585 int_values[1] = status.rx_count;
2586 keys[2] = "rstp_uptime";
2587 int_values[2] = status.uptime;
2588 keys[3] = "rstp_error_count";
2589 int_values[3] = status.error_count;
2590 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2591 ARRAY_SIZE(int_values));
2595 port_refresh_bond_status(struct port *port, bool force_update)
2597 struct eth_addr mac;
2599 /* Return if port is not a bond */
2600 if (list_is_singleton(&port->ifaces)) {
2604 if (bond_get_changed_active_slave(port->name, &mac, force_update)) {
2608 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2609 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2615 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2617 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2621 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2623 bool enable = enable_system_stats(cfg);
2625 system_stats_enable(enable);
2627 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2632 run_system_stats(void)
2634 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2637 stats = system_stats_run();
2639 struct ovsdb_idl_txn *txn;
2640 struct ovsdb_datum datum;
2642 txn = ovsdb_idl_txn_create(idl);
2643 ovsdb_datum_from_smap(&datum, stats);
2644 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2646 ovsdb_idl_txn_commit(txn);
2647 ovsdb_idl_txn_destroy(txn);
2654 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2657 case OFPCR12_ROLE_EQUAL:
2659 case OFPCR12_ROLE_MASTER:
2661 case OFPCR12_ROLE_SLAVE:
2663 case OFPCR12_ROLE_NOCHANGE:
2665 return "*** INVALID ROLE ***";
2670 refresh_controller_status(void)
2674 const struct ovsrec_controller *cfg;
2678 /* Accumulate status for controllers on all bridges. */
2679 HMAP_FOR_EACH (br, node, &all_bridges) {
2680 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2683 /* Update each controller in the database with current status. */
2684 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2685 struct ofproto_controller_info *cinfo =
2686 shash_find_data(&info, cfg->target);
2689 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2690 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2692 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2694 ovsrec_controller_set_is_connected(cfg, false);
2695 ovsrec_controller_set_role(cfg, NULL);
2696 ovsrec_controller_set_status(cfg, NULL);
2700 ofproto_free_ofproto_controller_info(&info);
2703 /* Update interface and mirror statistics if necessary. */
2705 run_stats_update(void)
2707 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2714 /* Statistics update interval should always be greater than or equal to
2716 stats_interval = MAX(smap_get_int(&cfg->other_config,
2717 "stats-update-interval",
2719 if (stats_timer_interval != stats_interval) {
2720 stats_timer_interval = stats_interval;
2721 stats_timer = LLONG_MIN;
2724 if (time_msec() >= stats_timer) {
2725 enum ovsdb_idl_txn_status status;
2727 /* Rate limit the update. Do not start a new update if the
2728 * previous one is not done. */
2732 stats_txn = ovsdb_idl_txn_create(idl);
2733 HMAP_FOR_EACH (br, node, &all_bridges) {
2737 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2738 struct iface *iface;
2740 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2741 iface_refresh_stats(iface);
2743 port_refresh_stp_stats(port);
2745 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2746 mirror_refresh_stats(m);
2749 refresh_controller_status();
2752 status = ovsdb_idl_txn_commit(stats_txn);
2753 if (status != TXN_INCOMPLETE) {
2754 stats_timer = time_msec() + stats_timer_interval;
2755 ovsdb_idl_txn_destroy(stats_txn);
2762 stats_update_wait(void)
2764 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2765 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2767 ovsdb_idl_txn_wait(stats_txn);
2769 poll_timer_wait_until(stats_timer);
2773 /* Update bridge/port/interface status if necessary. */
2775 run_status_update(void)
2780 /* Rate limit the update. Do not start a new update if the
2781 * previous one is not done. */
2782 seq = seq_read(connectivity_seq_get());
2783 if (seq != connectivity_seqno || status_txn_try_again) {
2786 connectivity_seqno = seq;
2787 status_txn = ovsdb_idl_txn_create(idl);
2788 HMAP_FOR_EACH (br, node, &all_bridges) {
2791 br_refresh_stp_status(br);
2792 br_refresh_rstp_status(br);
2793 br_refresh_datapath_info(br);
2794 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2795 struct iface *iface;
2797 port_refresh_stp_status(port);
2798 port_refresh_rstp_status(port);
2799 port_refresh_bond_status(port, status_txn_try_again);
2800 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2801 iface_refresh_netdev_status(iface);
2802 iface_refresh_ofproto_status(iface);
2809 /* Commit the transaction and get the status. If the transaction finishes,
2810 * then destroy the transaction. Otherwise, keep it so that we can check
2811 * progress the next time that this function is called. */
2813 enum ovsdb_idl_txn_status status;
2815 status = ovsdb_idl_txn_commit(status_txn);
2816 if (status != TXN_INCOMPLETE) {
2817 ovsdb_idl_txn_destroy(status_txn);
2820 /* Sets the 'status_txn_try_again' if the transaction fails. */
2821 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2822 status_txn_try_again = false;
2824 status_txn_try_again = true;
2829 /* Refresh AA port status if necessary. */
2830 if (time_msec() >= aa_refresh_timer) {
2833 HMAP_FOR_EACH (br, node, &all_bridges) {
2834 if (bridge_aa_need_refresh(br)) {
2835 struct ovsdb_idl_txn *txn;
2837 txn = ovsdb_idl_txn_create(idl);
2838 bridge_aa_refresh_queued(br);
2839 ovsdb_idl_txn_commit(txn);
2840 ovsdb_idl_txn_destroy(txn);
2844 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2849 status_update_wait(void)
2851 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2852 * transaction to complete. If the status update to database needs to be
2853 * run again (transaction fails), registers a timeout in
2854 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2855 * sequence number. */
2857 ovsdb_idl_txn_wait(status_txn);
2858 } else if (status_txn_try_again) {
2859 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2861 seq_wait(connectivity_seq_get(), connectivity_seqno);
2872 /* Let each datapath type do the work that it needs to do. */
2874 ofproto_enumerate_types(&types);
2875 SSET_FOR_EACH (type, &types) {
2876 ofproto_type_run(type);
2878 sset_destroy(&types);
2880 /* Let each bridge do the work that it needs to do. */
2881 HMAP_FOR_EACH (br, node, &all_bridges) {
2882 ofproto_run(br->ofproto);
2889 static struct ovsrec_open_vswitch null_cfg;
2890 const struct ovsrec_open_vswitch *cfg;
2892 bool vlan_splinters_changed;
2894 ovsrec_open_vswitch_init(&null_cfg);
2900 if (ovsdb_idl_is_lock_contended(idl)) {
2901 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2902 struct bridge *br, *next_br;
2904 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2905 "disabling this process (pid %ld) until it goes away",
2906 (long int) getpid());
2908 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2911 /* Since we will not be running system_stats_run() in this process
2912 * with the current situation of multiple ovs-vswitchd daemons,
2913 * disable system stats collection. */
2914 system_stats_enable(false);
2916 } else if (!ovsdb_idl_has_lock(idl)
2917 || !ovsdb_idl_has_ever_connected(idl)) {
2918 /* Returns if not holding the lock or not done retrieving db
2922 cfg = ovsrec_open_vswitch_first(idl);
2924 /* Initialize the ofproto library. This only needs to run once, but
2925 * it must be done after the configuration is set. If the
2926 * initialization has already occurred, bridge_init_ofproto()
2927 * returns immediately. */
2928 bridge_init_ofproto(cfg);
2930 /* Once the value of flow-restore-wait is false, we no longer should
2931 * check its value from the database. */
2932 if (cfg && ofproto_get_flow_restore_wait()) {
2933 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2934 "flow-restore-wait", false));
2939 /* Re-configure SSL. We do this on every trip through the main loop,
2940 * instead of just when the database changes, because the contents of the
2941 * key and certificate files can change without the database changing.
2943 * We do this before bridge_reconfigure() because that function might
2944 * initiate SSL connections and thus requires SSL to be configured. */
2945 if (cfg && cfg->ssl) {
2946 const struct ovsrec_ssl *ssl = cfg->ssl;
2948 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2949 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2952 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2953 * usage has changed. */
2954 vlan_splinters_changed = false;
2955 if (vlan_splinters_enabled_anywhere) {
2958 HMAP_FOR_EACH (br, node, &all_bridges) {
2959 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2960 vlan_splinters_changed = true;
2966 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed
2967 || ifaces_changed) {
2968 struct ovsdb_idl_txn *txn;
2970 ifaces_changed = false;
2972 idl_seqno = ovsdb_idl_get_seqno(idl);
2973 txn = ovsdb_idl_txn_create(idl);
2974 bridge_reconfigure(cfg ? cfg : &null_cfg);
2977 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2978 discover_types(cfg);
2981 /* If we are completing our initial configuration for this run
2982 * of ovs-vswitchd, then keep the transaction around to monitor
2983 * it for completion. */
2984 if (initial_config_done) {
2985 /* Always sets the 'status_txn_try_again' to check again,
2986 * in case that this transaction fails. */
2987 status_txn_try_again = true;
2988 ovsdb_idl_txn_commit(txn);
2989 ovsdb_idl_txn_destroy(txn);
2991 initial_config_done = true;
2992 daemonize_txn = txn;
2996 if (daemonize_txn) {
2997 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2998 if (status != TXN_INCOMPLETE) {
2999 ovsdb_idl_txn_destroy(daemonize_txn);
3000 daemonize_txn = NULL;
3002 /* ovs-vswitchd has completed initialization, so allow the
3003 * process that forked us to exit successfully. */
3004 daemonize_complete();
3006 vlog_enable_async();
3008 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3013 run_status_update();
3023 ovsdb_idl_wait(idl);
3024 if (daemonize_txn) {
3025 ovsdb_idl_txn_wait(daemonize_txn);
3029 if (ifaces_changed) {
3030 poll_immediate_wake();
3034 ofproto_enumerate_types(&types);
3035 SSET_FOR_EACH (type, &types) {
3036 ofproto_type_wait(type);
3038 sset_destroy(&types);
3040 if (!hmap_is_empty(&all_bridges)) {
3043 HMAP_FOR_EACH (br, node, &all_bridges) {
3044 ofproto_wait(br->ofproto);
3046 stats_update_wait();
3047 status_update_wait();
3050 system_stats_wait();
3053 /* Adds some memory usage statistics for bridges into 'usage', for use with
3054 * memory_report(). */
3056 bridge_get_memory_usage(struct simap *usage)
3063 ofproto_enumerate_types(&types);
3064 SSET_FOR_EACH (type, &types) {
3065 ofproto_type_get_memory_usage(type, usage);
3067 sset_destroy(&types);
3069 HMAP_FOR_EACH (br, node, &all_bridges) {
3070 ofproto_get_memory_usage(br->ofproto, usage);
3074 /* QoS unixctl user interface functions. */
3076 struct qos_unixctl_show_cbdata {
3078 struct iface *iface;
3082 qos_unixctl_show_queue(unsigned int queue_id,
3083 const struct smap *details,
3084 struct iface *iface,
3087 struct netdev_queue_stats stats;
3088 struct smap_node *node;
3091 ds_put_cstr(ds, "\n");
3093 ds_put_format(ds, "Queue %u:\n", queue_id);
3095 ds_put_cstr(ds, "Default:\n");
3098 SMAP_FOR_EACH (node, details) {
3099 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3102 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3104 if (stats.tx_packets != UINT64_MAX) {
3105 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3108 if (stats.tx_bytes != UINT64_MAX) {
3109 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3112 if (stats.tx_errors != UINT64_MAX) {
3113 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3116 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3117 queue_id, ovs_strerror(error));
3122 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3123 const char *argv[], void *aux OVS_UNUSED)
3125 struct ds ds = DS_EMPTY_INITIALIZER;
3126 struct smap smap = SMAP_INITIALIZER(&smap);
3127 struct iface *iface;
3129 struct smap_node *node;
3131 iface = iface_find(argv[1]);
3133 unixctl_command_reply_error(conn, "no such interface");
3137 netdev_get_qos(iface->netdev, &type, &smap);
3139 if (*type != '\0') {
3140 struct netdev_queue_dump dump;
3141 struct smap details;
3142 unsigned int queue_id;
3144 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3146 SMAP_FOR_EACH (node, &smap) {
3147 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3150 smap_init(&details);
3151 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3152 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3154 smap_destroy(&details);
3156 unixctl_command_reply(conn, ds_cstr(&ds));
3158 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3159 unixctl_command_reply_error(conn, ds_cstr(&ds));
3162 smap_destroy(&smap);
3166 /* Bridge reconfiguration functions. */
3168 bridge_create(const struct ovsrec_bridge *br_cfg)
3172 ovs_assert(!bridge_lookup(br_cfg->name));
3173 br = xzalloc(sizeof *br);
3175 br->name = xstrdup(br_cfg->name);
3176 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3179 /* Derive the default Ethernet address from the bridge's UUID. This should
3180 * be unique and it will be stable between ovs-vswitchd runs. */
3181 memcpy(&br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3182 eth_addr_mark_random(&br->default_ea);
3184 hmap_init(&br->ports);
3185 hmap_init(&br->ifaces);
3186 hmap_init(&br->iface_by_name);
3187 hmap_init(&br->mirrors);
3189 hmap_init(&br->mappings);
3190 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3194 bridge_destroy(struct bridge *br)
3197 struct mirror *mirror, *next_mirror;
3198 struct port *port, *next_port;
3200 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3203 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3204 mirror_destroy(mirror);
3207 hmap_remove(&all_bridges, &br->node);
3208 ofproto_destroy(br->ofproto);
3209 hmap_destroy(&br->ifaces);
3210 hmap_destroy(&br->ports);
3211 hmap_destroy(&br->iface_by_name);
3212 hmap_destroy(&br->mirrors);
3213 hmap_destroy(&br->mappings);
3220 static struct bridge *
3221 bridge_lookup(const char *name)
3225 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3226 if (!strcmp(br->name, name)) {
3233 /* Handle requests for a listing of all flows known by the OpenFlow
3234 * stack, including those normally hidden. */
3236 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3237 const char *argv[], void *aux OVS_UNUSED)
3242 br = bridge_lookup(argv[1]);
3244 unixctl_command_reply_error(conn, "Unknown bridge");
3249 ofproto_get_all_flows(br->ofproto, &results);
3251 unixctl_command_reply(conn, ds_cstr(&results));
3252 ds_destroy(&results);
3255 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3256 * connections and reconnect. If BRIDGE is not specified, then all bridges
3257 * drop their controller connections and reconnect. */
3259 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3260 const char *argv[], void *aux OVS_UNUSED)
3264 br = bridge_lookup(argv[1]);
3266 unixctl_command_reply_error(conn, "Unknown bridge");
3269 ofproto_reconnect_controllers(br->ofproto);
3271 HMAP_FOR_EACH (br, node, &all_bridges) {
3272 ofproto_reconnect_controllers(br->ofproto);
3275 unixctl_command_reply(conn, NULL);
3279 bridge_get_controllers(const struct bridge *br,
3280 struct ovsrec_controller ***controllersp)
3282 struct ovsrec_controller **controllers;
3283 size_t n_controllers;
3285 controllers = br->cfg->controller;
3286 n_controllers = br->cfg->n_controller;
3288 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3294 *controllersp = controllers;
3296 return n_controllers;
3300 bridge_collect_wanted_ports(struct bridge *br,
3301 const unsigned long int *splinter_vlans,
3302 struct shash *wanted_ports)
3306 shash_init(wanted_ports);
3308 for (i = 0; i < br->cfg->n_ports; i++) {
3309 const char *name = br->cfg->ports[i]->name;
3310 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3311 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3315 if (bridge_get_controllers(br, NULL)
3316 && !shash_find(wanted_ports, br->name)) {
3317 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3318 br->name, br->name);
3320 ovsrec_interface_init(&br->synth_local_iface);
3321 ovsrec_port_init(&br->synth_local_port);
3323 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3324 br->synth_local_port.n_interfaces = 1;
3325 br->synth_local_port.name = br->name;
3327 br->synth_local_iface.name = br->name;
3328 br->synth_local_iface.type = "internal";
3330 br->synth_local_ifacep = &br->synth_local_iface;
3332 shash_add(wanted_ports, br->name, &br->synth_local_port);
3335 if (splinter_vlans) {
3336 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3340 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3341 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3342 * 'br' needs to complete its configuration. */
3344 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3346 struct shash_node *port_node;
3347 struct port *port, *next;
3349 /* Get rid of deleted ports.
3350 * Get rid of deleted interfaces on ports that still exist. */
3351 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3352 port->cfg = shash_find_data(wanted_ports, port->name);
3356 port_del_ifaces(port);
3360 /* Update iface->cfg and iface->type in interfaces that still exist. */
3361 SHASH_FOR_EACH (port_node, wanted_ports) {
3362 const struct ovsrec_port *port = port_node->data;
3365 for (i = 0; i < port->n_interfaces; i++) {
3366 const struct ovsrec_interface *cfg = port->interfaces[i];
3367 struct iface *iface = iface_lookup(br, cfg->name);
3368 const char *type = iface_get_type(cfg, br->cfg);
3373 } else if (!strcmp(type, "null")) {
3374 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3375 " may be removed in February 2013. Please email"
3376 " dev@openvswitch.org with concerns.",
3379 /* We will add new interfaces later. */
3385 /* Initializes 'oc' appropriately as a management service controller for
3388 * The caller must free oc->target when it is no longer needed. */
3390 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3391 struct ofproto_controller *oc)
3393 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3394 oc->max_backoff = 0;
3395 oc->probe_interval = 60;
3396 oc->band = OFPROTO_OUT_OF_BAND;
3398 oc->burst_limit = 0;
3399 oc->enable_async_msgs = true;
3403 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3405 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3406 struct ofproto_controller *oc)
3410 oc->target = c->target;
3411 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3412 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3413 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3414 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3415 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3416 oc->burst_limit = (c->controller_burst_limit
3417 ? *c->controller_burst_limit : 0);
3418 oc->enable_async_msgs = (!c->enable_async_messages
3419 || *c->enable_async_messages);
3420 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3421 if (dscp < 0 || dscp > 63) {
3422 dscp = DSCP_DEFAULT;
3427 /* Configures the IP stack for 'br''s local interface properly according to the
3428 * configuration in 'c'. */
3430 bridge_configure_local_iface_netdev(struct bridge *br,
3431 struct ovsrec_controller *c)
3433 struct netdev *netdev;
3434 struct in_addr mask, gateway;
3436 struct iface *local_iface;
3439 /* If there's no local interface or no IP address, give up. */
3440 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3441 if (!local_iface || !c->local_ip
3442 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3446 /* Bring up the local interface. */
3447 netdev = local_iface->netdev;
3448 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3450 /* Configure the IP address and netmask. */
3451 if (!c->local_netmask
3452 || !inet_pton(AF_INET, c->local_netmask, &mask)
3454 mask.s_addr = guess_netmask(ip.s_addr);
3456 if (!netdev_set_in4(netdev, ip, mask)) {
3457 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3458 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3461 /* Configure the default gateway. */
3462 if (c->local_gateway
3463 && inet_pton(AF_INET, c->local_gateway, &gateway)
3464 && gateway.s_addr) {
3465 if (!netdev_add_router(netdev, gateway)) {
3466 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3467 br->name, IP_ARGS(gateway.s_addr));
3472 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3473 * in either string are treated as equal to any number of slashes in the other,
3474 * e.g. "x///y" is equal to "x/y".
3476 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3477 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3478 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3479 * 'b' against a prefix of 'a'.
3482 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3484 const char *b_start = b;
3486 if (b - b_start >= b_stoplen) {
3488 } else if (*a != *b) {
3490 } else if (*a == '/') {
3491 a += strspn(a, "/");
3492 b += strspn(b, "/");
3493 } else if (*a == '\0') {
3503 bridge_configure_remotes(struct bridge *br,
3504 const struct sockaddr_in *managers, size_t n_managers)
3506 bool disable_in_band;
3508 struct ovsrec_controller **controllers;
3509 size_t n_controllers;
3511 enum ofproto_fail_mode fail_mode;
3513 struct ofproto_controller *ocs;
3517 /* Check if we should disable in-band control on this bridge. */
3518 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3521 /* Set OpenFlow queue ID for in-band control. */
3522 ofproto_set_in_band_queue(br->ofproto,
3523 smap_get_int(&br->cfg->other_config,
3524 "in-band-queue", -1));
3526 if (disable_in_band) {
3527 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3529 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3532 n_controllers = bridge_get_controllers(br, &controllers);
3534 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3537 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3538 for (i = 0; i < n_controllers; i++) {
3539 struct ovsrec_controller *c = controllers[i];
3541 if (!strncmp(c->target, "punix:", 6)
3542 || !strncmp(c->target, "unix:", 5)) {
3543 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3546 if (!strncmp(c->target, "unix:", 5)) {
3547 /* Connect to a listening socket */
3548 whitelist = xasprintf("unix:%s/", ovs_rundir());
3549 if (strchr(c->target, '/') &&
3550 !equal_pathnames(c->target, whitelist,
3551 strlen(whitelist))) {
3552 /* Absolute path specified, but not in ovs_rundir */
3553 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3554 "controller \"%s\" due to possibility for "
3555 "remote exploit. Instead, specify socket "
3556 "in whitelisted \"%s\" or connect to "
3557 "\"unix:%s/%s.mgmt\" (which is always "
3558 "available without special configuration).",
3559 br->name, c->target, whitelist,
3560 ovs_rundir(), br->name);
3565 whitelist = xasprintf("punix:%s/%s.",
3566 ovs_rundir(), br->name);
3567 if (!equal_pathnames(c->target, whitelist, strlen(whitelist))
3568 || strchr(c->target + strlen(whitelist), '/')) {
3569 /* Prevent remote ovsdb-server users from accessing
3570 * arbitrary Unix domain sockets and overwriting arbitrary
3572 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3573 "controller \"%s\" due to possibility of "
3574 "overwriting local files. Instead, specify "
3575 "path in whitelisted format \"%s*\" or "
3576 "connect to \"unix:%s/%s.mgmt\" (which is "
3577 "always available without special "
3579 br->name, c->target, whitelist,
3580 ovs_rundir(), br->name);
3589 bridge_configure_local_iface_netdev(br, c);
3590 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3591 if (disable_in_band) {
3592 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3597 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3598 bridge_get_allowed_versions(br));
3599 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3602 /* Set the fail-mode. */
3603 fail_mode = !br->cfg->fail_mode
3604 || !strcmp(br->cfg->fail_mode, "standalone")
3605 ? OFPROTO_FAIL_STANDALONE
3606 : OFPROTO_FAIL_SECURE;
3607 ofproto_set_fail_mode(br->ofproto, fail_mode);
3609 /* Configure OpenFlow controller connection snooping. */
3610 if (!ofproto_has_snoops(br->ofproto)) {
3614 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3615 ovs_rundir(), br->name));
3616 ofproto_set_snoops(br->ofproto, &snoops);
3617 sset_destroy(&snoops);
3622 bridge_configure_tables(struct bridge *br)
3624 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3628 n_tables = ofproto_get_n_tables(br->ofproto);
3630 for (i = 0; i < n_tables; i++) {
3631 struct ofproto_table_settings s;
3632 bool use_default_prefixes = true;
3635 s.max_flows = UINT_MAX;
3637 s.enable_eviction = false;
3639 s.n_prefix_fields = 0;
3640 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3642 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3643 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3646 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3647 s.max_flows = *cfg->flow_limit;
3650 s.enable_eviction = (cfg->overflow_policy
3651 && !strcmp(cfg->overflow_policy, "evict"));
3652 if (cfg->n_groups) {
3653 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3654 for (k = 0; k < cfg->n_groups; k++) {
3655 const char *string = cfg->groups[k];
3658 msg = mf_parse_subfield__(&s.groups[k], &string);
3660 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3661 "'groups' (%s)", br->name, i, msg);
3663 } else if (*string) {
3664 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3665 "element '%s' contains trailing garbage",
3666 br->name, i, cfg->groups[k]);
3673 /* Prefix lookup fields. */
3674 s.n_prefix_fields = 0;
3675 for (k = 0; k < cfg->n_prefixes; k++) {
3676 const char *name = cfg->prefixes[k];
3677 const struct mf_field *mf;
3679 if (strcmp(name, "none") == 0) {
3680 use_default_prefixes = false;
3681 s.n_prefix_fields = 0;
3684 mf = mf_from_name(name);
3686 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3690 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3691 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3692 "%s", br->name, name);
3695 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3696 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3697 "field not used: %s", br->name, name);
3700 use_default_prefixes = false;
3701 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3704 if (use_default_prefixes) {
3705 /* Use default values. */
3706 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3707 memcpy(s.prefix_fields, default_prefix_fields,
3708 sizeof default_prefix_fields);
3711 struct ds ds = DS_EMPTY_INITIALIZER;
3712 for (k = 0; k < s.n_prefix_fields; k++) {
3714 ds_put_char(&ds, ',');
3716 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3718 if (s.n_prefix_fields == 0) {
3719 ds_put_cstr(&ds, "none");
3721 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3722 br->name, i, ds_cstr(&ds));
3726 ofproto_configure_table(br->ofproto, i, &s);
3730 for (; j < br->cfg->n_flow_tables; j++) {
3731 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3732 "%"PRId64" not supported by this datapath", br->name,
3733 br->cfg->key_flow_tables[j]);
3738 bridge_configure_dp_desc(struct bridge *br)
3740 ofproto_set_dp_desc(br->ofproto,
3741 smap_get(&br->cfg->other_config, "dp-desc"));
3744 static struct aa_mapping *
3745 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3747 struct aa_mapping *m;
3749 HMAP_FOR_EACH_IN_BUCKET (m,
3751 hash_bytes(&isid, sizeof isid, 0),
3753 if (isid == m->isid) {
3760 static struct aa_mapping *
3761 bridge_aa_mapping_create(struct bridge *br,
3765 struct aa_mapping *m;
3767 m = xzalloc(sizeof *m);
3771 m->br_name = xstrdup(br->name);
3772 hmap_insert(&br->mappings,
3774 hash_bytes(&isid, sizeof isid, 0));
3780 bridge_aa_mapping_destroy(struct aa_mapping *m)
3783 struct bridge *br = m->bridge;
3786 ofproto_aa_mapping_unregister(br->ofproto, m);
3789 hmap_remove(&br->mappings, &m->hmap_node);
3798 bridge_aa_mapping_configure(struct aa_mapping *m)
3800 struct aa_mapping_settings s;
3806 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3812 bridge_configure_aa(struct bridge *br)
3814 const struct ovsdb_datum *mc;
3815 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3816 struct aa_settings aa_s;
3817 struct aa_mapping *m, *next;
3821 ofproto_set_aa(br->ofproto, NULL, NULL);
3825 memset(&aa_s, 0, sizeof aa_s);
3826 aa_s.system_description = auto_attach->system_description;
3827 aa_s.system_name = auto_attach->system_name;
3828 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3830 mc = ovsrec_autoattach_get_mappings(auto_attach,
3832 OVSDB_TYPE_INTEGER);
3833 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3834 union ovsdb_atom atom;
3836 atom.integer = m->isid;
3837 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3838 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3840 bridge_aa_mapping_destroy(m);
3844 /* Add new mappings and reconfigure existing ones. */
3845 for (i = 0; i < auto_attach->n_mappings; ++i) {
3846 struct aa_mapping *m =
3847 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3850 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3851 auto_attach->key_mappings[i],
3852 auto_attach->value_mappings[i]);
3853 m = bridge_aa_mapping_create(br,
3854 auto_attach->key_mappings[i],
3855 auto_attach->value_mappings[i]);
3857 if (!bridge_aa_mapping_configure(m)) {
3858 bridge_aa_mapping_destroy(m);
3865 bridge_aa_need_refresh(struct bridge *br)
3867 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3871 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3873 int64_t *trunks = NULL;
3875 bool found = false, reconfigure = false;
3877 for (i = 0; i < port->cfg->n_trunks; i++) {
3878 if (port->cfg->trunks[i] == m->vlan) {
3885 case BRIDGE_AA_VLAN_OPER_ADD:
3887 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3889 for (i = 0; i < port->cfg->n_trunks; i++) {
3890 trunks[i] = port->cfg->trunks[i];
3892 trunks[i++] = m->vlan;
3898 case BRIDGE_AA_VLAN_OPER_REMOVE:
3902 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3904 for (i = 0; i < port->cfg->n_trunks; i++) {
3905 if (port->cfg->trunks[i] != m->vlan) {
3906 trunks[j++] = port->cfg->trunks[i];
3915 case BRIDGE_AA_VLAN_OPER_UNDEF:
3917 VLOG_WARN("unrecognized operation %u", m->oper);
3922 /* VLAN switching under trunk mode cause the trunk port to switch all
3923 * VLANs, see ovs-vswitchd.conf.db
3926 static char *vlan_mode_access = "access";
3927 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3931 static char *vlan_mode_trunk = "trunk";
3932 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3935 ovsrec_port_set_trunks(port->cfg, trunks, i);
3937 /* Force reconfigure of the port. */
3938 port_configure(port);
3943 bridge_aa_refresh_queued(struct bridge *br)
3945 struct ovs_list *list = xmalloc(sizeof *list);
3946 struct bridge_aa_vlan *node, *next;
3949 ofproto_aa_vlan_get_queued(br->ofproto, list);
3951 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3954 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3957 port = port_lookup(br, node->port_name);
3959 bridge_aa_update_trunks(port, node);
3962 list_remove(&node->list_node);
3963 free(node->port_name);
3971 /* Port functions. */
3973 static struct port *
3974 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3978 port = xzalloc(sizeof *port);
3980 port->name = xstrdup(cfg->name);
3982 list_init(&port->ifaces);
3984 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3988 /* Deletes interfaces from 'port' that are no longer configured for it. */
3990 port_del_ifaces(struct port *port)
3992 struct iface *iface, *next;
3993 struct sset new_ifaces;
3996 /* Collect list of new interfaces. */
3997 sset_init(&new_ifaces);
3998 for (i = 0; i < port->cfg->n_interfaces; i++) {
3999 const char *name = port->cfg->interfaces[i]->name;
4000 const char *type = port->cfg->interfaces[i]->type;
4001 if (strcmp(type, "null")) {
4002 sset_add(&new_ifaces, name);
4006 /* Get rid of deleted interfaces. */
4007 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4008 if (!sset_contains(&new_ifaces, iface->name)) {
4009 iface_destroy(iface);
4013 sset_destroy(&new_ifaces);
4017 port_destroy(struct port *port)
4020 struct bridge *br = port->bridge;
4021 struct iface *iface, *next;
4024 ofproto_bundle_unregister(br->ofproto, port);
4027 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4028 iface_destroy__(iface);
4031 hmap_remove(&br->ports, &port->hmap_node);
4037 static struct port *
4038 port_lookup(const struct bridge *br, const char *name)
4042 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4044 if (!strcmp(port->name, name)) {
4052 enable_lacp(struct port *port, bool *activep)
4054 if (!port->cfg->lacp) {
4055 /* XXX when LACP implementation has been sufficiently tested, enable by
4056 * default and make active on bonded ports. */
4058 } else if (!strcmp(port->cfg->lacp, "off")) {
4060 } else if (!strcmp(port->cfg->lacp, "active")) {
4063 } else if (!strcmp(port->cfg->lacp, "passive")) {
4067 VLOG_WARN("port %s: unknown LACP mode %s",
4068 port->name, port->cfg->lacp);
4073 static struct lacp_settings *
4074 port_configure_lacp(struct port *port, struct lacp_settings *s)
4076 const char *lacp_time, *system_id;
4079 if (!enable_lacp(port, &s->active)) {
4083 s->name = port->name;
4085 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4087 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4088 ETH_ADDR_SCAN_ARGS(s->id))) {
4089 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4090 " address.", port->name, system_id);
4094 s->id = port->bridge->ea;
4097 if (eth_addr_is_zero(s->id)) {
4098 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4102 /* Prefer bondable links if unspecified. */
4103 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4105 s->priority = (priority > 0 && priority <= UINT16_MAX
4107 : UINT16_MAX - !list_is_short(&port->ifaces));
4109 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4110 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4112 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4113 "lacp-fallback-ab", false);
4119 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4121 int priority, portid, key;
4123 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4124 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4126 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4128 if (portid <= 0 || portid > UINT16_MAX) {
4129 portid = ofp_to_u16(iface->ofp_port);
4132 if (priority <= 0 || priority > UINT16_MAX) {
4133 priority = UINT16_MAX;
4136 if (key < 0 || key > UINT16_MAX) {
4140 s->name = iface->name;
4142 s->priority = priority;
4147 port_configure_bond(struct port *port, struct bond_settings *s)
4149 const char *detect_s;
4150 struct iface *iface;
4152 int miimon_interval;
4154 s->name = port->name;
4156 if (port->cfg->bond_mode) {
4157 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4158 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4159 port->name, port->cfg->bond_mode,
4160 bond_mode_to_string(s->balance));
4163 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4165 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4166 * active-backup. At some point we should remove this warning. */
4167 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4168 " in previous versions, the default bond_mode was"
4169 " balance-slb", port->name,
4170 bond_mode_to_string(s->balance));
4172 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4173 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4174 "please use another bond type or disable flood_vlans",
4178 miimon_interval = smap_get_int(&port->cfg->other_config,
4179 "bond-miimon-interval", 0);
4180 if (miimon_interval <= 0) {
4181 miimon_interval = 200;
4184 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4185 if (!detect_s || !strcmp(detect_s, "carrier")) {
4186 miimon_interval = 0;
4187 } else if (strcmp(detect_s, "miimon")) {
4188 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4189 "defaulting to carrier", port->name, detect_s);
4190 miimon_interval = 0;
4193 s->up_delay = MAX(0, port->cfg->bond_updelay);
4194 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4195 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4196 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4197 "bond-rebalance-interval", 10000);
4198 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4199 s->rebalance_interval = 1000;
4202 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4203 "lacp-fallback-ab", false);
4205 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4206 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4209 mac_s = port->cfg->bond_active_slave;
4210 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4211 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4212 /* OVSDB did not store the last active interface */
4213 s->active_slave_mac = eth_addr_zero;
4217 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4218 * instead of obtaining it from the database. */
4220 port_is_synthetic(const struct port *port)
4222 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4225 /* Interface functions. */
4228 iface_is_internal(const struct ovsrec_interface *iface,
4229 const struct ovsrec_bridge *br)
4231 /* The local port and "internal" ports are always "internal". */
4232 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4235 /* Returns the correct network device type for interface 'iface' in bridge
4238 iface_get_type(const struct ovsrec_interface *iface,
4239 const struct ovsrec_bridge *br)
4243 /* The local port always has type "internal". Other ports take
4244 * their type from the database and default to "system" if none is
4246 if (iface_is_internal(iface, br)) {
4249 type = iface->type[0] ? iface->type : "system";
4252 return ofproto_port_open_type(br->datapath_type, type);
4256 iface_destroy__(struct iface *iface)
4259 struct port *port = iface->port;
4260 struct bridge *br = port->bridge;
4262 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4263 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4266 if (iface->ofp_port != OFPP_NONE) {
4267 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4270 list_remove(&iface->port_elem);
4271 hmap_remove(&br->iface_by_name, &iface->name_node);
4273 /* The user is changing configuration here, so netdev_remove needs to be
4274 * used as opposed to netdev_close */
4275 netdev_remove(iface->netdev);
4283 iface_destroy(struct iface *iface)
4286 struct port *port = iface->port;
4288 iface_destroy__(iface);
4289 if (list_is_empty(&port->ifaces)) {
4295 static struct iface *
4296 iface_lookup(const struct bridge *br, const char *name)
4298 struct iface *iface;
4300 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4301 &br->iface_by_name) {
4302 if (!strcmp(iface->name, name)) {
4310 static struct iface *
4311 iface_find(const char *name)
4313 const struct bridge *br;
4315 HMAP_FOR_EACH (br, node, &all_bridges) {
4316 struct iface *iface = iface_lookup(br, name);
4325 static struct iface *
4326 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4328 struct iface *iface;
4330 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4332 if (iface->ofp_port == ofp_port) {
4339 /* Set Ethernet address of 'iface', if one is specified in the configuration
4342 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4344 struct eth_addr ea, *mac = NULL;
4345 struct iface *hw_addr_iface;
4347 if (strcmp(iface->type, "internal")) {
4351 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, &ea)) {
4353 } else if (port->cfg->fake_bridge) {
4354 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4355 find_local_hw_addr(br, &ea, port, &hw_addr_iface);
4360 if (iface->ofp_port == OFPP_LOCAL) {
4361 VLOG_ERR("interface %s: ignoring mac in Interface record "
4362 "(use Bridge record to set local port's mac)",
4364 } else if (eth_addr_is_multicast(*mac)) {
4365 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4368 int error = netdev_set_etheraddr(iface->netdev, *mac);
4370 VLOG_ERR("interface %s: setting MAC failed (%s)",
4371 iface->name, ovs_strerror(error));
4377 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4379 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4381 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4382 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4383 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4387 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4388 * sets the "ofport" field to -1.
4390 * This is appropriate when 'if_cfg''s interface cannot be created or is
4391 * otherwise invalid. */
4393 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4395 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4396 iface_set_ofport(if_cfg, OFPP_NONE);
4397 ovsrec_interface_set_error(if_cfg, errp);
4398 ovsrec_interface_set_status(if_cfg, NULL);
4399 ovsrec_interface_set_admin_state(if_cfg, NULL);
4400 ovsrec_interface_set_duplex(if_cfg, NULL);
4401 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4402 ovsrec_interface_set_link_state(if_cfg, NULL);
4403 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4404 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4405 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4406 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4407 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4408 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4409 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4410 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4415 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4417 union ovsdb_atom atom;
4419 atom.integer = target;
4420 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4424 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4426 struct ofpbuf queues_buf;
4428 ofpbuf_init(&queues_buf, 0);
4430 if (!qos || qos->type[0] == '\0') {
4431 netdev_set_qos(iface->netdev, NULL, NULL);
4433 const struct ovsdb_datum *queues;
4434 struct netdev_queue_dump dump;
4435 unsigned int queue_id;
4436 struct smap details;
4440 /* Configure top-level Qos for 'iface'. */
4441 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4443 /* Deconfigure queues that were deleted. */
4444 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4446 smap_init(&details);
4447 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4448 if (!queue_ids_include(queues, queue_id)) {
4449 netdev_delete_queue(iface->netdev, queue_id);
4452 smap_destroy(&details);
4454 /* Configure queues for 'iface'. */
4456 for (i = 0; i < qos->n_queues; i++) {
4457 const struct ovsrec_queue *queue = qos->value_queues[i];
4458 unsigned int queue_id = qos->key_queues[i];
4460 if (queue_id == 0) {
4464 if (queue->n_dscp == 1) {
4465 struct ofproto_port_queue *port_queue;
4467 port_queue = ofpbuf_put_uninit(&queues_buf,
4468 sizeof *port_queue);
4469 port_queue->queue = queue_id;
4470 port_queue->dscp = queue->dscp[0];
4473 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4476 struct smap details;
4478 smap_init(&details);
4479 netdev_set_queue(iface->netdev, 0, &details);
4480 smap_destroy(&details);
4484 if (iface->ofp_port != OFPP_NONE) {
4485 const struct ofproto_port_queue *port_queues = queues_buf.data;
4486 size_t n_queues = queues_buf.size / sizeof *port_queues;
4488 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4489 port_queues, n_queues);
4492 netdev_set_policing(iface->netdev,
4493 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4494 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4496 ofpbuf_uninit(&queues_buf);
4500 iface_configure_cfm(struct iface *iface)
4502 const struct ovsrec_interface *cfg = iface->cfg;
4503 const char *opstate_str;
4504 const char *cfm_ccm_vlan;
4505 struct cfm_settings s;
4506 struct smap netdev_args;
4508 if (!cfg->n_cfm_mpid) {
4509 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4513 s.check_tnl_key = false;
4514 smap_init(&netdev_args);
4515 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4516 const char *key = smap_get(&netdev_args, "key");
4517 const char *in_key = smap_get(&netdev_args, "in_key");
4519 s.check_tnl_key = (key && !strcmp(key, "flow"))
4520 || (in_key && !strcmp(in_key, "flow"));
4522 smap_destroy(&netdev_args);
4524 s.mpid = *cfg->cfm_mpid;
4525 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4526 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4527 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4529 if (s.interval <= 0) {
4533 if (!cfm_ccm_vlan) {
4535 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4536 s.ccm_vlan = CFM_RANDOM_VLAN;
4538 s.ccm_vlan = atoi(cfm_ccm_vlan);
4539 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4544 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4546 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4548 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4549 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4551 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4554 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4555 * instead of obtaining it from the database. */
4557 iface_is_synthetic(const struct iface *iface)
4559 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4563 iface_validate_ofport__(size_t n, int64_t *ofport)
4565 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4566 ? u16_to_ofp(*ofport)
4571 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4573 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4577 iface_pick_ofport(const struct ovsrec_interface *cfg)
4579 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4580 return (requested_ofport != OFPP_NONE
4582 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4585 /* Port mirroring. */
4587 static struct mirror *
4588 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4592 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4593 if (uuid_equals(uuid, &m->uuid)) {
4601 bridge_configure_mirrors(struct bridge *br)
4603 const struct ovsdb_datum *mc;
4604 unsigned long *flood_vlans;
4605 struct mirror *m, *next;
4608 /* Get rid of deleted mirrors. */
4609 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4610 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4611 union ovsdb_atom atom;
4613 atom.uuid = m->uuid;
4614 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4619 /* Add new mirrors and reconfigure existing ones. */
4620 for (i = 0; i < br->cfg->n_mirrors; i++) {
4621 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4622 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4624 m = mirror_create(br, cfg);
4627 if (!mirror_configure(m)) {
4632 /* Update flooded vlans (for RSPAN). */
4633 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4634 br->cfg->n_flood_vlans);
4635 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4636 bitmap_free(flood_vlans);
4639 static struct mirror *
4640 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4644 m = xzalloc(sizeof *m);
4645 m->uuid = cfg->header_.uuid;
4646 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4648 m->name = xstrdup(cfg->name);
4654 mirror_destroy(struct mirror *m)
4657 struct bridge *br = m->bridge;
4660 ofproto_mirror_unregister(br->ofproto, m);
4663 hmap_remove(&br->mirrors, &m->hmap_node);
4670 mirror_collect_ports(struct mirror *m,
4671 struct ovsrec_port **in_ports, int n_in_ports,
4672 void ***out_portsp, size_t *n_out_portsp)
4674 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4675 size_t n_out_ports = 0;
4678 for (i = 0; i < n_in_ports; i++) {
4679 const char *name = in_ports[i]->name;
4680 struct port *port = port_lookup(m->bridge, name);
4682 out_ports[n_out_ports++] = port;
4684 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4685 "port %s", m->bridge->name, m->name, name);
4688 *out_portsp = out_ports;
4689 *n_out_portsp = n_out_ports;
4693 mirror_configure(struct mirror *m)
4695 const struct ovsrec_mirror *cfg = m->cfg;
4696 struct ofproto_mirror_settings s;
4699 if (strcmp(cfg->name, m->name)) {
4701 m->name = xstrdup(cfg->name);
4705 /* Get output port or VLAN. */
4706 if (cfg->output_port) {
4707 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4708 if (!s.out_bundle) {
4709 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4710 m->bridge->name, m->name);
4713 s.out_vlan = UINT16_MAX;
4715 if (cfg->output_vlan) {
4716 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4717 "output vlan; ignoring output vlan",
4718 m->bridge->name, m->name);
4720 } else if (cfg->output_vlan) {
4721 /* The database should prevent invalid VLAN values. */
4722 s.out_bundle = NULL;
4723 s.out_vlan = *cfg->output_vlan;
4725 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4726 m->bridge->name, m->name);
4730 /* Get port selection. */
4731 if (cfg->select_all) {
4732 size_t n_ports = hmap_count(&m->bridge->ports);
4733 void **ports = xmalloc(n_ports * sizeof *ports);
4738 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4748 /* Get ports, dropping ports that don't exist.
4749 * The IDL ensures that there are no duplicates. */
4750 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4751 &s.srcs, &s.n_srcs);
4752 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4753 &s.dsts, &s.n_dsts);
4756 /* Get VLAN selection. */
4757 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4760 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4763 if (s.srcs != s.dsts) {
4772 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4774 * This is deprecated. It is only for compatibility with broken device drivers
4775 * in old versions of Linux that do not properly support VLANs when VLAN
4776 * devices are not used. When broken device drivers are no longer in
4777 * widespread use, we will delete these interfaces. */
4779 static struct ovsrec_port **recs;
4780 static size_t n_recs, allocated_recs;
4782 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4783 * splinters are reconfigured. */
4785 register_rec(struct ovsrec_port *rec)
4787 if (n_recs >= allocated_recs) {
4788 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4790 recs[n_recs++] = rec;
4793 /* Frees all of the ports registered with register_reg(). */
4795 free_registered_recs(void)
4799 for (i = 0; i < n_recs; i++) {
4800 struct ovsrec_port *port = recs[i];
4803 for (j = 0; j < port->n_interfaces; j++) {
4804 struct ovsrec_interface *iface = port->interfaces[j];
4809 smap_destroy(&port->other_config);
4810 free(port->interfaces);
4818 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4821 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4823 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4827 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4830 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4831 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4832 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4835 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4836 * use, returns NULL.
4838 * Updates 'vlan_splinters_enabled_anywhere'. */
4839 static unsigned long int *
4840 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4842 unsigned long int *splinter_vlans;
4843 struct sset splinter_ifaces;
4844 const char *real_dev_name;
4845 struct shash *real_devs;
4846 struct shash_node *node;
4850 /* Free space allocated for synthesized ports and interfaces, since we're
4851 * in the process of reconstructing all of them. */
4852 free_registered_recs();
4854 splinter_vlans = bitmap_allocate(4096);
4855 sset_init(&splinter_ifaces);
4856 vlan_splinters_enabled_anywhere = false;
4857 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4858 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4861 for (j = 0; j < br_cfg->n_ports; j++) {
4862 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4865 for (k = 0; k < port_cfg->n_interfaces; k++) {
4866 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4868 if (vlan_splinters_is_enabled(iface_cfg)) {
4869 vlan_splinters_enabled_anywhere = true;
4870 sset_add(&splinter_ifaces, iface_cfg->name);
4871 vlan_bitmap_from_array__(port_cfg->trunks,
4877 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4878 bitmap_set1(splinter_vlans, *port_cfg->tag);
4883 if (!vlan_splinters_enabled_anywhere) {
4884 free(splinter_vlans);
4885 sset_destroy(&splinter_ifaces);
4889 HMAP_FOR_EACH (br, node, &all_bridges) {
4891 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4895 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4896 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4897 * device to be created for it. */
4898 bitmap_set0(splinter_vlans, 0);
4899 bitmap_set0(splinter_vlans, 4095);
4901 /* Delete all VLAN devices that we don't need. */
4903 real_devs = vlandev_get_real_devs();
4904 SHASH_FOR_EACH (node, real_devs) {
4905 const struct vlan_real_dev *real_dev = node->data;
4906 const struct vlan_dev *vlan_dev;
4907 bool real_dev_has_splinters;
4909 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4911 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4912 if (!real_dev_has_splinters
4913 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4914 struct netdev *netdev;
4916 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4917 if (!netdev_get_in4(netdev, NULL, NULL) ||
4918 !netdev_get_in6(netdev, NULL)) {
4919 /* It has an IP address configured, so we don't own
4920 * it. Don't delete it. */
4922 vlandev_del(vlan_dev->name);
4924 netdev_close(netdev);
4931 /* Add all VLAN devices that we need. */
4932 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4935 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4936 if (!vlandev_get_name(real_dev_name, vid)) {
4937 vlandev_add(real_dev_name, vid);
4944 sset_destroy(&splinter_ifaces);
4946 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4947 free(splinter_vlans);
4950 return splinter_vlans;
4953 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4956 configure_splinter_port(struct port *port)
4958 struct ofproto *ofproto = port->bridge->ofproto;
4959 ofp_port_t realdev_ofp_port;
4960 const char *realdev_name;
4961 struct iface *vlandev, *realdev;
4963 ofproto_bundle_unregister(port->bridge->ofproto, port);
4965 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4968 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4969 realdev = iface_lookup(port->bridge, realdev_name);
4970 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4972 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4976 static struct ovsrec_port *
4977 synthesize_splinter_port(const char *real_dev_name,
4978 const char *vlan_dev_name, int vid)
4980 struct ovsrec_interface *iface;
4981 struct ovsrec_port *port;
4983 iface = xmalloc(sizeof *iface);
4984 ovsrec_interface_init(iface);
4985 iface->name = xstrdup(vlan_dev_name);
4986 iface->type = "system";
4988 port = xmalloc(sizeof *port);
4989 ovsrec_port_init(port);
4990 port->interfaces = xmemdup(&iface, sizeof iface);
4991 port->n_interfaces = 1;
4992 port->name = xstrdup(vlan_dev_name);
4993 port->vlan_mode = "splinter";
4994 port->tag = xmalloc(sizeof *port->tag);
4997 smap_add(&port->other_config, "realdev", real_dev_name);
5003 /* For each interface with 'br' that has VLAN splinters enabled, adds a
5004 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
5005 * 1-bit in the 'splinter_vlans' bitmap. */
5007 add_vlan_splinter_ports(struct bridge *br,
5008 const unsigned long int *splinter_vlans,
5009 struct shash *ports)
5013 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
5014 * we're modifying 'ports'. */
5015 for (i = 0; i < br->cfg->n_ports; i++) {
5016 const char *name = br->cfg->ports[i]->name;
5017 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
5020 for (j = 0; j < port_cfg->n_interfaces; j++) {
5021 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
5023 if (vlan_splinters_is_enabled(iface_cfg)) {
5024 const char *real_dev_name;
5027 real_dev_name = iface_cfg->name;
5028 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
5029 const char *vlan_dev_name;
5031 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5033 && !shash_find(ports, vlan_dev_name)) {
5034 shash_add(ports, vlan_dev_name,
5035 synthesize_splinter_port(
5036 real_dev_name, vlan_dev_name, vid));
5045 mirror_refresh_stats(struct mirror *m)
5047 struct ofproto *ofproto = m->bridge->ofproto;
5048 uint64_t tx_packets, tx_bytes;
5049 const char *keys[2];
5051 size_t stat_cnt = 0;
5053 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5054 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5058 if (tx_packets != UINT64_MAX) {
5059 keys[stat_cnt] = "tx_packets";
5060 values[stat_cnt] = tx_packets;
5063 if (tx_bytes != UINT64_MAX) {
5064 keys[stat_cnt] = "tx_bytes";
5065 values[stat_cnt] = tx_bytes;
5069 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5073 * Add registered netdev and dpif types to ovsdb to allow external
5074 * applications to query the capabilities of the Open vSwitch instance
5075 * running on the node.
5078 discover_types(const struct ovsrec_open_vswitch *cfg)
5082 /* Datapath types. */
5084 dp_enumerate_types(&types);
5085 const char **datapath_types = sset_array(&types);
5086 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5087 sset_count(&types));
5088 free(datapath_types);
5089 sset_destroy(&types);
5093 netdev_enumerate_types(&types);
5094 const char **iface_types = sset_array(&types);
5095 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5097 sset_destroy(&types);