1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 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 *, bool del);
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) {
503 bridge_destroy(br, false);
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_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
587 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
588 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
590 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
591 * to 'ovs_cfg', with only very minimal configuration otherwise.
593 * This is mostly an update to bridge data structures. Nothing is pushed
594 * down to ofproto or lower layers. */
595 add_del_bridges(ovs_cfg);
596 splinter_vlans = collect_splinter_vlans(ovs_cfg);
597 HMAP_FOR_EACH (br, node, &all_bridges) {
598 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
599 bridge_del_ports(br, &br->wanted_ports);
601 free(splinter_vlans);
603 /* Start pushing configuration changes down to the ofproto layer:
605 * - Delete ofprotos that are no longer configured.
607 * - Delete ports that are no longer configured.
609 * - Reconfigure existing ports to their desired configurations, or
610 * delete them if not possible.
612 * We have to do all the deletions before we can do any additions, because
613 * the ports to be added might require resources that will be freed up by
614 * deletions (they might especially overlap in name). */
615 bridge_delete_ofprotos();
616 HMAP_FOR_EACH (br, node, &all_bridges) {
618 bridge_delete_or_reconfigure_ports(br);
622 /* Finish pushing configuration changes to the ofproto layer:
624 * - Create ofprotos that are missing.
626 * - Add ports that are missing. */
627 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
631 error = ofproto_create(br->name, br->type, &br->ofproto);
633 VLOG_ERR("failed to create bridge %s: %s", br->name,
634 ovs_strerror(error));
635 shash_destroy(&br->wanted_ports);
636 bridge_destroy(br, true);
638 /* Trigger storing datapath version. */
639 seq_change(connectivity_seq_get());
643 HMAP_FOR_EACH (br, node, &all_bridges) {
644 bridge_add_ports(br, &br->wanted_ports);
645 shash_destroy(&br->wanted_ports);
648 reconfigure_system_stats(ovs_cfg);
650 /* Complete the configuration. */
651 sflow_bridge_number = 0;
652 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
653 HMAP_FOR_EACH (br, node, &all_bridges) {
656 /* We need the datapath ID early to allow LACP ports to use it as the
657 * default system ID. */
658 bridge_configure_datapath_id(br);
660 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
663 port_configure(port);
665 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
666 iface_set_ofport(iface->cfg, iface->ofp_port);
667 /* Clear eventual previous errors */
668 ovsrec_interface_set_error(iface->cfg, NULL);
669 iface_configure_cfm(iface);
670 iface_configure_qos(iface, port->cfg->qos);
671 iface_set_mac(br, port, iface);
672 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
674 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
678 bridge_configure_mirrors(br);
679 bridge_configure_forward_bpdu(br);
680 bridge_configure_mac_table(br);
681 bridge_configure_mcast_snooping(br);
682 bridge_configure_remotes(br, managers, n_managers);
683 bridge_configure_netflow(br);
684 bridge_configure_sflow(br, &sflow_bridge_number);
685 bridge_configure_ipfix(br);
686 bridge_configure_spanning_tree(br);
687 bridge_configure_tables(br);
688 bridge_configure_dp_desc(br);
689 bridge_configure_aa(br);
693 /* The ofproto-dpif provider does some final reconfiguration in its
694 * ->type_run() function. We have to call it before notifying the database
695 * client that reconfiguration is complete, otherwise there is a very
696 * narrow race window in which e.g. ofproto/trace will not recognize the
697 * new configuration (sometimes this causes unit test failures). */
701 /* Delete ofprotos which aren't configured or have the wrong type. Create
702 * ofprotos which don't exist but need to. */
704 bridge_delete_ofprotos(void)
711 /* Delete ofprotos with no bridge or with the wrong type. */
714 ofproto_enumerate_types(&types);
715 SSET_FOR_EACH (type, &types) {
718 ofproto_enumerate_names(type, &names);
719 SSET_FOR_EACH (name, &names) {
720 br = bridge_lookup(name);
721 if (!br || strcmp(type, br->type)) {
722 ofproto_delete(name, type);
726 sset_destroy(&names);
727 sset_destroy(&types);
731 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
733 if (*n >= *allocated) {
734 ports = x2nrealloc(ports, allocated, sizeof *ports);
736 ports[(*n)++] = port;
741 bridge_delete_or_reconfigure_ports(struct bridge *br)
743 struct ofproto_port ofproto_port;
744 struct ofproto_port_dump dump;
746 struct sset ofproto_ports;
747 struct port *port, *port_next;
749 /* List of "ofp_port"s to delete. We make a list instead of deleting them
750 * right away because ofproto implementations aren't necessarily able to
751 * iterate through a changing list of ports in an entirely robust way. */
758 sset_init(&ofproto_ports);
760 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
761 * that are not configured in the database. (This commonly happens when
762 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
764 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
765 * that have the wrong OpenFlow port number (and arrange to add them back
766 * with the correct OpenFlow port number). */
767 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
768 ofp_port_t requested_ofp_port;
771 sset_add(&ofproto_ports, ofproto_port.name);
773 iface = iface_lookup(br, ofproto_port.name);
775 /* No such iface is configured, so we should delete this
778 * As a corner case exception, keep the port if it's a bond fake
780 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
781 && !strcmp(ofproto_port.type, "internal")) {
787 if (strcmp(ofproto_port.type, iface->type)
788 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
789 /* The interface is the wrong type or can't be configured.
794 /* If the requested OpenFlow port for 'iface' changed, and it's not
795 * already the correct port, then we might want to temporarily delete
796 * this interface, so we can add it back again with the new OpenFlow
798 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
799 if (iface->ofp_port != OFPP_LOCAL &&
800 requested_ofp_port != OFPP_NONE &&
801 requested_ofp_port != iface->ofp_port) {
802 ofp_port_t victim_request;
803 struct iface *victim;
805 /* Check for an existing OpenFlow port currently occupying
806 * 'iface''s requested port number. If there isn't one, then
807 * delete this port. Otherwise we need to consider further. */
808 victim = iface_from_ofp_port(br, requested_ofp_port);
813 /* 'victim' is a port currently using 'iface''s requested port
814 * number. Unless 'victim' specifically requested that port
815 * number, too, then we can delete both 'iface' and 'victim'
816 * temporarily. (We'll add both of them back again later with new
817 * OpenFlow port numbers.)
819 * If 'victim' did request port number 'requested_ofp_port', just
820 * like 'iface', then that's a configuration inconsistency that we
821 * can't resolve. We might as well let it keep its current port
823 victim_request = iface_get_requested_ofp_port(victim->cfg);
824 if (victim_request != requested_ofp_port) {
825 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
826 iface_destroy(victim);
835 iface_destroy(iface);
836 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
838 for (i = 0; i < n; i++) {
839 ofproto_port_del(br->ofproto, del[i]);
843 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
844 * that we didn't see when we iterated through the datapath, i.e. ports
845 * that disappeared underneath use. This is an unusual situation, but it
846 * can happen in some cases:
848 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
849 * idea but could happen).
851 * - The port represented a device that disappeared, e.g. a tuntap
852 * device destroyed via "tunctl -d", a physical Ethernet device
853 * whose module was just unloaded via "rmmod", or a virtual NIC for a
854 * VM whose VM was just terminated. */
855 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
856 struct iface *iface, *iface_next;
858 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
859 if (!sset_contains(&ofproto_ports, iface->name)) {
860 iface_destroy__(iface);
864 if (list_is_empty(&port->ifaces)) {
868 sset_destroy(&ofproto_ports);
872 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
873 bool with_requested_port)
875 struct shash_node *port_node;
877 SHASH_FOR_EACH (port_node, wanted_ports) {
878 const struct ovsrec_port *port_cfg = port_node->data;
881 for (i = 0; i < port_cfg->n_interfaces; i++) {
882 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
883 ofp_port_t requested_ofp_port;
885 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
886 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
887 struct iface *iface = iface_lookup(br, iface_cfg->name);
890 iface_create(br, iface_cfg, port_cfg);
898 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
900 /* First add interfaces that request a particular port number. */
901 bridge_add_ports__(br, wanted_ports, true);
903 /* Then add interfaces that want automatic port number assignment.
904 * We add these afterward to avoid accidentally taking a specifically
905 * requested port number. */
906 bridge_add_ports__(br, wanted_ports, false);
910 port_configure(struct port *port)
912 const struct ovsrec_port *cfg = port->cfg;
913 struct bond_settings bond_settings;
914 struct lacp_settings lacp_settings;
915 struct ofproto_bundle_settings s;
918 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
919 configure_splinter_port(port);
928 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
929 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
930 s.slaves[s.n_slaves++] = iface->ofp_port;
935 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
939 /* Get VLAN trunks. */
942 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
946 if (cfg->vlan_mode) {
947 if (!strcmp(cfg->vlan_mode, "access")) {
948 s.vlan_mode = PORT_VLAN_ACCESS;
949 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
950 s.vlan_mode = PORT_VLAN_TRUNK;
951 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
952 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
953 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
954 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
956 /* This "can't happen" because ovsdb-server should prevent it. */
957 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
958 "back to trunk mode", port->name, cfg->vlan_mode);
959 s.vlan_mode = PORT_VLAN_TRUNK;
963 s.vlan_mode = PORT_VLAN_ACCESS;
965 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
969 s.vlan_mode = PORT_VLAN_TRUNK;
972 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
975 /* Get LACP settings. */
976 s.lacp = port_configure_lacp(port, &lacp_settings);
980 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
981 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
982 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
985 s.lacp_slaves = NULL;
988 /* Get bond settings. */
989 if (s.n_slaves > 1) {
990 s.bond = &bond_settings;
991 port_configure_bond(port, &bond_settings);
994 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
995 netdev_set_miimon_interval(iface->netdev, 0);
1000 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1005 free(s.lacp_slaves);
1008 /* Pick local port hardware address and datapath ID for 'br'. */
1010 bridge_configure_datapath_id(struct bridge *br)
1014 struct iface *local_iface;
1015 struct iface *hw_addr_iface;
1018 bridge_pick_local_hw_addr(br, &ea, &hw_addr_iface);
1019 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1021 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1023 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1024 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1025 "Ethernet address: %s",
1026 br->name, ovs_strerror(error));
1031 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1032 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1033 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1034 ofproto_set_datapath_id(br->ofproto, dpid);
1037 dpid_string = xasprintf("%016"PRIx64, dpid);
1038 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1042 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1045 bridge_get_allowed_versions(struct bridge *br)
1047 if (!br->cfg->n_protocols) {
1051 return ofputil_versions_from_strings(br->cfg->protocols,
1052 br->cfg->n_protocols);
1055 /* Set NetFlow configuration on 'br'. */
1057 bridge_configure_netflow(struct bridge *br)
1059 struct ovsrec_netflow *cfg = br->cfg->netflow;
1060 struct netflow_options opts;
1063 ofproto_set_netflow(br->ofproto, NULL);
1067 memset(&opts, 0, sizeof opts);
1069 /* Get default NetFlow configuration from datapath.
1070 * Apply overrides from 'cfg'. */
1071 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1072 if (cfg->engine_type) {
1073 opts.engine_type = *cfg->engine_type;
1075 if (cfg->engine_id) {
1076 opts.engine_id = *cfg->engine_id;
1079 /* Configure active timeout interval. */
1080 opts.active_timeout = cfg->active_timeout;
1081 if (!opts.active_timeout) {
1082 opts.active_timeout = -1;
1083 } else if (opts.active_timeout < 0) {
1084 VLOG_WARN("bridge %s: active timeout interval set to negative "
1085 "value, using default instead (%d seconds)", br->name,
1086 NF_ACTIVE_TIMEOUT_DEFAULT);
1087 opts.active_timeout = -1;
1090 /* Add engine ID to interface number to disambiguate bridgs? */
1091 opts.add_id_to_iface = cfg->add_id_to_interface;
1092 if (opts.add_id_to_iface) {
1093 if (opts.engine_id > 0x7f) {
1094 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1095 "another vswitch, choose an engine id less than 128",
1098 if (hmap_count(&br->ports) > 508) {
1099 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1100 "another port when more than 508 ports are used",
1106 sset_init(&opts.collectors);
1107 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1110 if (ofproto_set_netflow(br->ofproto, &opts)) {
1111 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1113 sset_destroy(&opts.collectors);
1116 /* Set sFlow configuration on 'br'. */
1118 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1120 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1121 struct ovsrec_controller **controllers;
1122 struct ofproto_sflow_options oso;
1123 size_t n_controllers;
1127 ofproto_set_sflow(br->ofproto, NULL);
1131 memset(&oso, 0, sizeof oso);
1133 sset_init(&oso.targets);
1134 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1136 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1137 if (cfg->sampling) {
1138 oso.sampling_rate = *cfg->sampling;
1141 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1143 oso.polling_interval = *cfg->polling;
1146 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1148 oso.header_len = *cfg->header;
1151 oso.sub_id = (*sflow_bridge_number)++;
1152 oso.agent_device = cfg->agent;
1154 oso.control_ip = NULL;
1155 n_controllers = bridge_get_controllers(br, &controllers);
1156 for (i = 0; i < n_controllers; i++) {
1157 if (controllers[i]->local_ip) {
1158 oso.control_ip = controllers[i]->local_ip;
1162 ofproto_set_sflow(br->ofproto, &oso);
1164 sset_destroy(&oso.targets);
1167 /* Returns whether a IPFIX row is valid. */
1169 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1171 return ipfix && ipfix->n_targets > 0;
1174 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1176 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1177 const struct bridge *br)
1179 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1182 /* Set IPFIX configuration on 'br'. */
1184 bridge_configure_ipfix(struct bridge *br)
1186 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1187 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1188 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1189 struct ofproto_ipfix_bridge_exporter_options be_opts;
1190 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1191 size_t n_fe_opts = 0;
1193 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1194 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1199 if (!valid_be_cfg && n_fe_opts == 0) {
1200 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1205 memset(&be_opts, 0, sizeof be_opts);
1207 sset_init(&be_opts.targets);
1208 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1210 if (be_cfg->sampling) {
1211 be_opts.sampling_rate = *be_cfg->sampling;
1213 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1215 if (be_cfg->obs_domain_id) {
1216 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1218 if (be_cfg->obs_point_id) {
1219 be_opts.obs_point_id = *be_cfg->obs_point_id;
1221 if (be_cfg->cache_active_timeout) {
1222 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1224 if (be_cfg->cache_max_flows) {
1225 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1228 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1229 "enable-tunnel-sampling", true);
1231 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1232 "enable-input-sampling", false);
1234 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1235 "enable-output-sampling", false);
1238 if (n_fe_opts > 0) {
1239 struct ofproto_ipfix_flow_exporter_options *opts;
1240 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1242 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1243 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1244 opts->collector_set_id = fe_cfg->id;
1245 sset_init(&opts->targets);
1246 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1247 fe_cfg->ipfix->n_targets);
1248 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1249 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1250 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1251 ? *fe_cfg->ipfix->cache_max_flows : 0;
1257 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1261 sset_destroy(&be_opts.targets);
1264 if (n_fe_opts > 0) {
1265 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1267 for (i = 0; i < n_fe_opts; i++) {
1268 sset_destroy(&opts->targets);
1276 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1277 struct ofproto_port_stp_settings *port_s,
1278 int *port_num_counter, unsigned long *port_num_bitmap)
1280 const char *config_str;
1281 struct iface *iface;
1283 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1284 port_s->enable = false;
1287 port_s->enable = true;
1290 /* STP over bonds is not supported. */
1291 if (!list_is_singleton(&port->ifaces)) {
1292 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1294 port_s->enable = false;
1298 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1300 /* Internal ports shouldn't participate in spanning tree, so
1302 if (!strcmp(iface->type, "internal")) {
1303 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1304 port_s->enable = false;
1308 /* STP on mirror output ports is not supported. */
1309 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1310 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1311 port_s->enable = false;
1315 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1317 unsigned long int port_num = strtoul(config_str, NULL, 0);
1318 int port_idx = port_num - 1;
1320 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1321 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1322 port_s->enable = false;
1326 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1327 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1328 port->name, port_num);
1329 port_s->enable = false;
1332 bitmap_set1(port_num_bitmap, port_idx);
1333 port_s->port_num = port_idx;
1335 if (*port_num_counter >= STP_MAX_PORTS) {
1336 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1337 port_s->enable = false;
1341 port_s->port_num = (*port_num_counter)++;
1344 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1346 port_s->path_cost = strtoul(config_str, NULL, 10);
1348 enum netdev_features current;
1351 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1352 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1353 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1356 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1358 port_s->priority = strtoul(config_str, NULL, 0);
1360 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1365 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1366 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1368 const char *config_str;
1369 struct iface *iface;
1371 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1372 port_s->enable = false;
1375 port_s->enable = true;
1378 /* RSTP over bonds is not supported. */
1379 if (!list_is_singleton(&port->ifaces)) {
1380 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1382 port_s->enable = false;
1386 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1388 /* Internal ports shouldn't participate in spanning tree, so
1390 if (!strcmp(iface->type, "internal")) {
1391 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1392 port_s->enable = false;
1396 /* RSTP on mirror output ports is not supported. */
1397 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1398 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1399 port_s->enable = false;
1403 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1405 unsigned long int port_num = strtoul(config_str, NULL, 0);
1406 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1407 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1408 port_s->enable = false;
1411 port_s->port_num = port_num;
1413 if (*port_num_counter >= RSTP_MAX_PORTS) {
1414 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1415 port_s->enable = false;
1418 /* If rstp-port-num is not specified, use 0.
1419 * rstp_port_set_port_number() will look for the first free one. */
1420 port_s->port_num = 0;
1423 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1425 port_s->path_cost = strtoul(config_str, NULL, 10);
1427 enum netdev_features current;
1430 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1431 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1432 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1435 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1437 port_s->priority = strtoul(config_str, NULL, 0);
1439 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1442 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1444 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1446 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1449 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1450 "rstp-admin-port-state", true);
1452 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1453 "rstp-port-admin-edge", false);
1454 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1455 "rstp-port-auto-edge", true);
1456 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1457 "rstp-port-mcheck", false);
1460 /* Set spanning tree configuration on 'br'. */
1462 bridge_configure_stp(struct bridge *br, bool enable_stp)
1465 ofproto_set_stp(br->ofproto, NULL);
1467 struct ofproto_stp_settings br_s;
1468 const char *config_str;
1470 int port_num_counter;
1471 unsigned long *port_num_bitmap;
1473 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1477 if (eth_addr_from_string(config_str, &ea)) {
1478 br_s.system_id = eth_addr_to_uint64(ea);
1480 br_s.system_id = eth_addr_to_uint64(br->ea);
1481 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1482 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1485 br_s.system_id = eth_addr_to_uint64(br->ea);
1488 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1490 br_s.priority = strtoul(config_str, NULL, 0);
1492 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1495 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1497 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1499 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1502 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1504 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1506 br_s.max_age = STP_DEFAULT_MAX_AGE;
1509 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1511 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1513 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1516 /* Configure STP on the bridge. */
1517 if (ofproto_set_stp(br->ofproto, &br_s)) {
1518 VLOG_ERR("bridge %s: could not enable STP", br->name);
1522 /* Users must either set the port number with the "stp-port-num"
1523 * configuration on all ports or none. If manual configuration
1524 * is not done, then we allocate them sequentially. */
1525 port_num_counter = 0;
1526 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1527 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1528 struct ofproto_port_stp_settings port_s;
1529 struct iface *iface;
1531 port_configure_stp(br->ofproto, port, &port_s,
1532 &port_num_counter, port_num_bitmap);
1534 /* As bonds are not supported, just apply configuration to
1535 * all interfaces. */
1536 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1537 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1539 VLOG_ERR("port %s: could not enable STP", port->name);
1545 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1546 && port_num_counter) {
1547 VLOG_ERR("bridge %s: must manually configure all STP port "
1548 "IDs or none, disabling", br->name);
1549 ofproto_set_stp(br->ofproto, NULL);
1551 bitmap_free(port_num_bitmap);
1556 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1559 ofproto_set_rstp(br->ofproto, NULL);
1561 struct ofproto_rstp_settings br_s;
1562 const char *config_str;
1564 int port_num_counter;
1566 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1570 if (eth_addr_from_string(config_str, &ea)) {
1571 br_s.address = eth_addr_to_uint64(ea);
1574 br_s.address = eth_addr_to_uint64(br->ea);
1575 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1576 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1580 br_s.address = eth_addr_to_uint64(br->ea);
1583 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1585 br_s.priority = strtoul(config_str, NULL, 0);
1587 br_s.priority = RSTP_DEFAULT_PRIORITY;
1590 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1592 br_s.ageing_time = strtoul(config_str, NULL, 0);
1594 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1597 config_str = smap_get(&br->cfg->other_config,
1598 "rstp-force-protocol-version");
1600 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1602 br_s.force_protocol_version = FPV_DEFAULT;
1605 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1607 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1609 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1612 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1614 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1616 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1619 config_str = smap_get(&br->cfg->other_config,
1620 "rstp-transmit-hold-count");
1622 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1624 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1627 /* Configure RSTP on the bridge. */
1628 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1629 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1633 port_num_counter = 0;
1634 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1635 struct ofproto_port_rstp_settings port_s;
1636 struct iface *iface;
1638 port_configure_rstp(br->ofproto, port, &port_s,
1641 /* As bonds are not supported, just apply configuration to
1642 * all interfaces. */
1643 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1644 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1646 VLOG_ERR("port %s: could not enable RSTP", port->name);
1655 bridge_configure_spanning_tree(struct bridge *br)
1657 bool enable_rstp = br->cfg->rstp_enable;
1658 bool enable_stp = br->cfg->stp_enable;
1660 if (enable_rstp && enable_stp) {
1661 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1662 "configured; enabling RSTP", br->name);
1666 bridge_configure_stp(br, enable_stp);
1667 bridge_configure_rstp(br, enable_rstp);
1671 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1673 const struct port *port = port_lookup(br, name);
1674 return port && port_is_bond_fake_iface(port);
1678 port_is_bond_fake_iface(const struct port *port)
1680 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1684 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1686 struct bridge *br, *next;
1687 struct shash_node *node;
1688 struct shash new_br;
1691 /* Collect new bridges' names and types. */
1692 shash_init(&new_br);
1693 for (i = 0; i < cfg->n_bridges; i++) {
1694 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1695 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1697 if (strchr(br_cfg->name, '/') || strchr(br_cfg->name, '\\')) {
1698 /* Prevent remote ovsdb-server users from accessing arbitrary
1699 * directories, e.g. consider a bridge named "../../../etc/".
1701 * Prohibiting "\" is only necessary on Windows but it's no great
1702 * loss elsewhere. */
1703 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1705 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1706 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1710 /* Get rid of deleted bridges or those whose types have changed.
1711 * Update 'cfg' of bridges that still exist. */
1712 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1713 br->cfg = shash_find_data(&new_br, br->name);
1714 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1715 br->cfg->datapath_type))) {
1716 bridge_destroy(br, true);
1720 /* Add new bridges. */
1721 SHASH_FOR_EACH(node, &new_br) {
1722 const struct ovsrec_bridge *br_cfg = node->data;
1723 struct bridge *br = bridge_lookup(br_cfg->name);
1725 bridge_create(br_cfg);
1729 shash_destroy(&new_br);
1732 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1733 * Returns 0 if successful, otherwise a positive errno value. */
1735 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1736 struct netdev *netdev, char **errp)
1738 return netdev_set_config(netdev, &iface_cfg->options, errp);
1741 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1742 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1744 * If successful, returns 0 and stores the network device in '*netdevp'. On
1745 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1747 iface_do_create(const struct bridge *br,
1748 const struct ovsrec_interface *iface_cfg,
1749 const struct ovsrec_port *port_cfg,
1750 ofp_port_t *ofp_portp, struct netdev **netdevp,
1753 struct netdev *netdev = NULL;
1756 if (netdev_is_reserved_name(iface_cfg->name)) {
1757 VLOG_WARN("could not create interface %s, name is reserved",
1763 error = netdev_open(iface_cfg->name,
1764 iface_get_type(iface_cfg, br->cfg), &netdev);
1766 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1767 iface_cfg->name, ovs_strerror(error));
1771 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1776 *ofp_portp = iface_pick_ofport(iface_cfg);
1777 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1782 VLOG_INFO("bridge %s: added interface %s on port %d",
1783 br->name, iface_cfg->name, *ofp_portp);
1785 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1786 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1794 netdev_close(netdev);
1798 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1799 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1800 * automatically allocated for the iface. Takes ownership of and
1801 * deallocates 'if_cfg'.
1803 * Return true if an iface is successfully created, false otherwise. */
1805 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1806 const struct ovsrec_port *port_cfg)
1808 struct netdev *netdev;
1809 struct iface *iface;
1810 ofp_port_t ofp_port;
1815 /* Do the bits that can fail up front. */
1816 ovs_assert(!iface_lookup(br, iface_cfg->name));
1817 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1819 iface_clear_db_record(iface_cfg, errp);
1824 /* Get or create the port structure. */
1825 port = port_lookup(br, port_cfg->name);
1827 port = port_create(br, port_cfg);
1830 /* Create the iface structure. */
1831 iface = xzalloc(sizeof *iface);
1832 list_push_back(&port->ifaces, &iface->port_elem);
1833 hmap_insert(&br->iface_by_name, &iface->name_node,
1834 hash_string(iface_cfg->name, 0));
1836 iface->name = xstrdup(iface_cfg->name);
1837 iface->ofp_port = ofp_port;
1838 iface->netdev = netdev;
1839 iface->type = iface_get_type(iface_cfg, br->cfg);
1840 iface->cfg = iface_cfg;
1841 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1842 hash_ofp_port(ofp_port));
1844 /* Populate initial status in database. */
1845 iface_refresh_stats(iface);
1846 iface_refresh_netdev_status(iface);
1848 /* Add bond fake iface if necessary. */
1849 if (port_is_bond_fake_iface(port)) {
1850 struct ofproto_port ofproto_port;
1852 if (ofproto_port_query_by_name(br->ofproto, port->name,
1854 struct netdev *netdev;
1857 error = netdev_open(port->name, "internal", &netdev);
1859 ofp_port_t fake_ofp_port = OFPP_NONE;
1860 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1861 netdev_close(netdev);
1863 VLOG_WARN("could not open network device %s (%s)",
1864 port->name, ovs_strerror(error));
1867 /* Already exists, nothing to do. */
1868 ofproto_port_destroy(&ofproto_port);
1875 /* Set forward BPDU option. */
1877 bridge_configure_forward_bpdu(struct bridge *br)
1879 ofproto_set_forward_bpdu(br->ofproto,
1880 smap_get_bool(&br->cfg->other_config,
1885 /* Set MAC learning table configuration for 'br'. */
1887 bridge_configure_mac_table(struct bridge *br)
1889 const char *idle_time_str;
1892 const char *mac_table_size_str;
1895 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1896 idle_time = (idle_time_str && atoi(idle_time_str)
1897 ? atoi(idle_time_str)
1898 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1900 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1901 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1902 ? atoi(mac_table_size_str)
1905 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1908 /* Set multicast snooping table configuration for 'br'. */
1910 bridge_configure_mcast_snooping(struct bridge *br)
1912 if (!br->cfg->mcast_snooping_enable) {
1913 ofproto_set_mcast_snooping(br->ofproto, NULL);
1916 struct ofproto_mcast_snooping_settings br_s;
1917 const char *idle_time_str;
1918 const char *max_entries_str;
1920 idle_time_str = smap_get(&br->cfg->other_config,
1921 "mcast-snooping-aging-time");
1922 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1923 ? atoi(idle_time_str)
1924 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1926 max_entries_str = smap_get(&br->cfg->other_config,
1927 "mcast-snooping-table-size");
1928 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1929 ? atoi(max_entries_str)
1930 : MCAST_DEFAULT_MAX_ENTRIES);
1932 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1933 "mcast-snooping-disable-flood-unregistered",
1936 /* Configure multicast snooping on the bridge */
1937 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1938 VLOG_ERR("bridge %s: could not enable multicast snooping",
1943 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1944 struct ofproto_mcast_snooping_port_settings port_s;
1945 port_s.flood = smap_get_bool(&port->cfg->other_config,
1946 "mcast-snooping-flood", false);
1947 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1948 "mcast-snooping-flood-reports", false);
1949 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1950 VLOG_ERR("port %s: could not configure mcast snooping",
1958 find_local_hw_addr(const struct bridge *br, struct eth_addr *ea,
1959 const struct port *fake_br, struct iface **hw_addr_iface)
1961 struct hmapx mirror_output_ports;
1963 bool found_addr = false;
1967 /* Mirror output ports don't participate in picking the local hardware
1968 * address. ofproto can't help us find out whether a given port is a
1969 * mirror output because we haven't configured mirrors yet, so we need to
1970 * accumulate them ourselves. */
1971 hmapx_init(&mirror_output_ports);
1972 for (i = 0; i < br->cfg->n_mirrors; i++) {
1973 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1974 if (m->output_port) {
1975 hmapx_add(&mirror_output_ports, m->output_port);
1979 /* Otherwise choose the minimum non-local MAC address among all of the
1981 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1982 struct eth_addr iface_ea;
1983 struct iface *candidate;
1984 struct iface *iface;
1986 /* Mirror output ports don't participate. */
1987 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1991 /* Choose the MAC address to represent the port. */
1993 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac,
1995 /* Find the interface with this Ethernet address (if any) so that
1996 * we can provide the correct devname to the caller. */
1997 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1998 struct eth_addr candidate_ea;
1999 if (!netdev_get_etheraddr(candidate->netdev, &candidate_ea)
2000 && eth_addr_equals(iface_ea, candidate_ea)) {
2005 /* Choose the interface whose MAC address will represent the port.
2006 * The Linux kernel bonding code always chooses the MAC address of
2007 * the first slave added to a bond, and the Fedora networking
2008 * scripts always add slaves to a bond in alphabetical order, so
2009 * for compatibility we choose the interface with the name that is
2010 * first in alphabetical order. */
2011 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2012 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2017 /* The local port doesn't count (since we're trying to choose its
2018 * MAC address anyway). */
2019 if (iface->ofp_port == OFPP_LOCAL) {
2023 /* For fake bridges we only choose from ports with the same tag */
2024 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2025 if (!port->cfg->tag) {
2028 if (*port->cfg->tag != *fake_br->cfg->tag) {
2034 error = netdev_get_etheraddr(iface->netdev, &iface_ea);
2040 /* Compare against our current choice. */
2041 if (!eth_addr_is_multicast(iface_ea) &&
2042 !eth_addr_is_local(iface_ea) &&
2043 !eth_addr_is_reserved(iface_ea) &&
2044 !eth_addr_is_zero(iface_ea) &&
2045 (!found_addr || eth_addr_compare_3way(iface_ea, *ea) < 0))
2048 *hw_addr_iface = iface;
2054 *ea = br->default_ea;
2055 *hw_addr_iface = NULL;
2058 hmapx_destroy(&mirror_output_ports);
2062 bridge_pick_local_hw_addr(struct bridge *br, struct eth_addr *ea,
2063 struct iface **hw_addr_iface)
2066 *hw_addr_iface = NULL;
2068 /* Did the user request a particular MAC? */
2069 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2070 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2071 if (eth_addr_is_multicast(*ea)) {
2072 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2073 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(*ea));
2074 } else if (eth_addr_is_zero(*ea)) {
2075 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2081 /* Find a local hw address */
2082 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2085 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2086 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2087 * an interface on 'br', then that interface must be passed in as
2088 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2089 * 'hw_addr_iface' must be passed in as a null pointer. */
2091 bridge_pick_datapath_id(struct bridge *br,
2092 const struct eth_addr bridge_ea,
2093 struct iface *hw_addr_iface)
2096 * The procedure for choosing a bridge MAC address will, in the most
2097 * ordinary case, also choose a unique MAC that we can use as a datapath
2098 * ID. In some special cases, though, multiple bridges will end up with
2099 * the same MAC address. This is OK for the bridges, but it will confuse
2100 * the OpenFlow controller, because each datapath needs a unique datapath
2103 * Datapath IDs must be unique. It is also very desirable that they be
2104 * stable from one run to the next, so that policy set on a datapath
2107 const char *datapath_id;
2110 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2111 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2115 if (!hw_addr_iface) {
2117 * A purely internal bridge, that is, one that has no non-virtual
2118 * network devices on it at all, is difficult because it has no
2119 * natural unique identifier at all.
2121 * When the host is a XenServer, we handle this case by hashing the
2122 * host's UUID with the name of the bridge. Names of bridges are
2123 * persistent across XenServer reboots, although they can be reused if
2124 * an internal network is destroyed and then a new one is later
2125 * created, so this is fairly effective.
2127 * When the host is not a XenServer, we punt by using a random MAC
2128 * address on each run.
2130 const char *host_uuid = xenserver_get_host_uuid();
2132 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2133 dpid = dpid_from_hash(combined, strlen(combined));
2139 return eth_addr_to_uint64(bridge_ea);
2143 dpid_from_hash(const void *data, size_t n)
2146 uint8_t bytes[SHA1_DIGEST_SIZE];
2150 sha1_bytes(data, n, hash.bytes);
2151 eth_addr_mark_random(&hash.ea);
2152 return eth_addr_to_uint64(hash.ea);
2156 iface_refresh_netdev_status(struct iface *iface)
2160 enum netdev_features current;
2161 enum netdev_flags flags;
2162 const char *link_state;
2163 struct eth_addr mac;
2164 int64_t bps, mtu_64, ifindex64, link_resets;
2167 if (iface_is_synthetic(iface)) {
2171 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2172 && !status_txn_try_again) {
2176 iface->change_seq = netdev_get_change_seq(iface->netdev);
2180 if (!netdev_get_status(iface->netdev, &smap)) {
2181 ovsrec_interface_set_status(iface->cfg, &smap);
2183 ovsrec_interface_set_status(iface->cfg, NULL);
2186 smap_destroy(&smap);
2188 error = netdev_get_flags(iface->netdev, &flags);
2190 const char *state = flags & NETDEV_UP ? "up" : "down";
2192 ovsrec_interface_set_admin_state(iface->cfg, state);
2194 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2197 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2198 ovsrec_interface_set_link_state(iface->cfg, link_state);
2200 link_resets = netdev_get_carrier_resets(iface->netdev);
2201 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2203 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2204 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2206 ovsrec_interface_set_duplex(iface->cfg,
2207 netdev_features_is_full_duplex(current)
2209 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2211 ovsrec_interface_set_duplex(iface->cfg, NULL);
2212 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2215 error = netdev_get_mtu(iface->netdev, &mtu);
2218 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2220 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2223 error = netdev_get_etheraddr(iface->netdev, &mac);
2225 char mac_string[ETH_ADDR_STRLEN + 1];
2227 snprintf(mac_string, sizeof mac_string,
2228 ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2229 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2231 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2234 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2235 * if there is no valid ifindex number. */
2236 ifindex64 = netdev_get_ifindex(iface->netdev);
2237 if (ifindex64 < 0) {
2240 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2244 iface_refresh_ofproto_status(struct iface *iface)
2248 if (iface_is_synthetic(iface)) {
2252 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2256 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2258 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2261 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2263 || status_txn_try_again) {
2264 iface_refresh_cfm_stats(iface);
2267 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2269 || status_txn_try_again) {
2273 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2274 iface->ofp_port, &smap);
2275 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2276 smap_destroy(&smap);
2280 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2283 iface_refresh_cfm_stats(struct iface *iface)
2285 const struct ovsrec_interface *cfg = iface->cfg;
2286 struct cfm_status status;
2289 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2290 iface->ofp_port, &status);
2292 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2293 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2294 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2295 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2296 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2297 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2299 const char *reasons[CFM_FAULT_N_REASONS];
2300 int64_t cfm_health = status.health;
2301 int64_t cfm_flap_count = status.flap_count;
2302 bool faulted = status.faults != 0;
2305 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2308 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2309 int reason = 1 << i;
2310 if (status.faults & reason) {
2311 reasons[j++] = cfm_fault_reason_to_str(reason);
2314 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2316 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2318 if (status.remote_opstate >= 0) {
2319 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2320 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2322 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2325 ovsrec_interface_set_cfm_remote_mpids(cfg,
2326 (const int64_t *)status.rmps,
2328 if (cfm_health >= 0) {
2329 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2331 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2339 iface_refresh_stats(struct iface *iface)
2341 #define IFACE_STATS \
2342 IFACE_STAT(rx_packets, "rx_packets") \
2343 IFACE_STAT(tx_packets, "tx_packets") \
2344 IFACE_STAT(rx_bytes, "rx_bytes") \
2345 IFACE_STAT(tx_bytes, "tx_bytes") \
2346 IFACE_STAT(rx_dropped, "rx_dropped") \
2347 IFACE_STAT(tx_dropped, "tx_dropped") \
2348 IFACE_STAT(rx_errors, "rx_errors") \
2349 IFACE_STAT(tx_errors, "tx_errors") \
2350 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2351 IFACE_STAT(rx_over_errors, "rx_over_err") \
2352 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2353 IFACE_STAT(collisions, "collisions")
2355 #define IFACE_STAT(MEMBER, NAME) + 1
2356 enum { N_IFACE_STATS = IFACE_STATS };
2358 int64_t values[N_IFACE_STATS];
2359 const char *keys[N_IFACE_STATS];
2362 struct netdev_stats stats;
2364 if (iface_is_synthetic(iface)) {
2368 /* Intentionally ignore return value, since errors will set 'stats' to
2369 * all-1s, and we will deal with that correctly below. */
2370 netdev_get_stats(iface->netdev, &stats);
2372 /* Copy statistics into keys[] and values[]. */
2374 #define IFACE_STAT(MEMBER, NAME) \
2375 if (stats.MEMBER != UINT64_MAX) { \
2377 values[n] = stats.MEMBER; \
2382 ovs_assert(n <= N_IFACE_STATS);
2384 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2389 br_refresh_datapath_info(struct bridge *br)
2391 const char *version;
2393 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2394 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2397 ovsrec_bridge_set_datapath_version(br->cfg,
2398 version ? version : "<unknown>");
2402 br_refresh_stp_status(struct bridge *br)
2404 struct smap smap = SMAP_INITIALIZER(&smap);
2405 struct ofproto *ofproto = br->ofproto;
2406 struct ofproto_stp_status status;
2408 if (ofproto_get_stp_status(ofproto, &status)) {
2412 if (!status.enabled) {
2413 ovsrec_bridge_set_status(br->cfg, NULL);
2417 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2418 STP_ID_ARGS(status.bridge_id));
2419 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2420 STP_ID_ARGS(status.designated_root));
2421 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2423 ovsrec_bridge_set_status(br->cfg, &smap);
2424 smap_destroy(&smap);
2428 port_refresh_stp_status(struct port *port)
2430 struct ofproto *ofproto = port->bridge->ofproto;
2431 struct iface *iface;
2432 struct ofproto_port_stp_status status;
2435 if (port_is_synthetic(port)) {
2439 /* STP doesn't currently support bonds. */
2440 if (!list_is_singleton(&port->ifaces)) {
2441 ovsrec_port_set_status(port->cfg, NULL);
2445 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2446 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2450 if (!status.enabled) {
2451 ovsrec_port_set_status(port->cfg, NULL);
2455 /* Set Status column. */
2457 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2458 smap_add(&smap, "stp_state", stp_state_name(status.state));
2459 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2460 smap_add(&smap, "stp_role", stp_role_name(status.role));
2461 ovsrec_port_set_status(port->cfg, &smap);
2462 smap_destroy(&smap);
2466 port_refresh_stp_stats(struct port *port)
2468 struct ofproto *ofproto = port->bridge->ofproto;
2469 struct iface *iface;
2470 struct ofproto_port_stp_stats stats;
2471 const char *keys[3];
2472 int64_t int_values[3];
2474 if (port_is_synthetic(port)) {
2478 /* STP doesn't currently support bonds. */
2479 if (!list_is_singleton(&port->ifaces)) {
2483 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2484 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2488 if (!stats.enabled) {
2489 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2493 /* Set Statistics column. */
2494 keys[0] = "stp_tx_count";
2495 int_values[0] = stats.tx_count;
2496 keys[1] = "stp_rx_count";
2497 int_values[1] = stats.rx_count;
2498 keys[2] = "stp_error_count";
2499 int_values[2] = stats.error_count;
2501 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2502 ARRAY_SIZE(int_values));
2506 br_refresh_rstp_status(struct bridge *br)
2508 struct smap smap = SMAP_INITIALIZER(&smap);
2509 struct ofproto *ofproto = br->ofproto;
2510 struct ofproto_rstp_status status;
2512 if (ofproto_get_rstp_status(ofproto, &status)) {
2515 if (!status.enabled) {
2516 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2519 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2520 RSTP_ID_ARGS(status.bridge_id));
2521 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2522 status.root_path_cost);
2523 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2524 RSTP_ID_ARGS(status.root_id));
2525 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2526 RSTP_ID_ARGS(status.designated_id));
2527 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2528 status.designated_port_id);
2529 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2530 status.bridge_port_id);
2531 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2532 smap_destroy(&smap);
2536 port_refresh_rstp_status(struct port *port)
2538 struct ofproto *ofproto = port->bridge->ofproto;
2539 struct iface *iface;
2540 struct ofproto_port_rstp_status status;
2541 const char *keys[4];
2542 int64_t int_values[4];
2545 if (port_is_synthetic(port)) {
2549 /* RSTP doesn't currently support bonds. */
2550 if (!list_is_singleton(&port->ifaces)) {
2551 ovsrec_port_set_rstp_status(port->cfg, NULL);
2555 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2556 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2560 if (!status.enabled) {
2561 ovsrec_port_set_rstp_status(port->cfg, NULL);
2562 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2565 /* Set Status column. */
2568 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2570 smap_add_format(&smap, "rstp_port_role", "%s",
2571 rstp_port_role_name(status.role));
2572 smap_add_format(&smap, "rstp_port_state", "%s",
2573 rstp_state_name(status.state));
2574 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2575 RSTP_ID_ARGS(status.designated_bridge_id));
2576 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2577 status.designated_port_id);
2578 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2579 status.designated_path_cost);
2581 ovsrec_port_set_rstp_status(port->cfg, &smap);
2582 smap_destroy(&smap);
2584 /* Set Statistics column. */
2585 keys[0] = "rstp_tx_count";
2586 int_values[0] = status.tx_count;
2587 keys[1] = "rstp_rx_count";
2588 int_values[1] = status.rx_count;
2589 keys[2] = "rstp_uptime";
2590 int_values[2] = status.uptime;
2591 keys[3] = "rstp_error_count";
2592 int_values[3] = status.error_count;
2593 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2594 ARRAY_SIZE(int_values));
2598 port_refresh_bond_status(struct port *port, bool force_update)
2600 struct eth_addr mac;
2602 /* Return if port is not a bond */
2603 if (list_is_singleton(&port->ifaces)) {
2607 if (bond_get_changed_active_slave(port->name, &mac, force_update)) {
2611 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2612 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2618 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2620 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2624 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2626 bool enable = enable_system_stats(cfg);
2628 system_stats_enable(enable);
2630 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2635 run_system_stats(void)
2637 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2640 stats = system_stats_run();
2642 struct ovsdb_idl_txn *txn;
2643 struct ovsdb_datum datum;
2645 txn = ovsdb_idl_txn_create(idl);
2646 ovsdb_datum_from_smap(&datum, stats);
2647 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2649 ovsdb_idl_txn_commit(txn);
2650 ovsdb_idl_txn_destroy(txn);
2657 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2660 case OFPCR12_ROLE_EQUAL:
2662 case OFPCR12_ROLE_MASTER:
2664 case OFPCR12_ROLE_SLAVE:
2666 case OFPCR12_ROLE_NOCHANGE:
2668 return "*** INVALID ROLE ***";
2673 refresh_controller_status(void)
2677 const struct ovsrec_controller *cfg;
2681 /* Accumulate status for controllers on all bridges. */
2682 HMAP_FOR_EACH (br, node, &all_bridges) {
2683 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2686 /* Update each controller in the database with current status. */
2687 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2688 struct ofproto_controller_info *cinfo =
2689 shash_find_data(&info, cfg->target);
2692 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2693 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2695 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2697 ovsrec_controller_set_is_connected(cfg, false);
2698 ovsrec_controller_set_role(cfg, NULL);
2699 ovsrec_controller_set_status(cfg, NULL);
2703 ofproto_free_ofproto_controller_info(&info);
2706 /* Update interface and mirror statistics if necessary. */
2708 run_stats_update(void)
2710 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2717 /* Statistics update interval should always be greater than or equal to
2719 stats_interval = MAX(smap_get_int(&cfg->other_config,
2720 "stats-update-interval",
2722 if (stats_timer_interval != stats_interval) {
2723 stats_timer_interval = stats_interval;
2724 stats_timer = LLONG_MIN;
2727 if (time_msec() >= stats_timer) {
2728 enum ovsdb_idl_txn_status status;
2730 /* Rate limit the update. Do not start a new update if the
2731 * previous one is not done. */
2735 stats_txn = ovsdb_idl_txn_create(idl);
2736 HMAP_FOR_EACH (br, node, &all_bridges) {
2740 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2741 struct iface *iface;
2743 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2744 iface_refresh_stats(iface);
2746 port_refresh_stp_stats(port);
2748 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2749 mirror_refresh_stats(m);
2752 refresh_controller_status();
2755 status = ovsdb_idl_txn_commit(stats_txn);
2756 if (status != TXN_INCOMPLETE) {
2757 stats_timer = time_msec() + stats_timer_interval;
2758 ovsdb_idl_txn_destroy(stats_txn);
2765 stats_update_wait(void)
2767 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2768 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2770 ovsdb_idl_txn_wait(stats_txn);
2772 poll_timer_wait_until(stats_timer);
2776 /* Update bridge/port/interface status if necessary. */
2778 run_status_update(void)
2783 /* Rate limit the update. Do not start a new update if the
2784 * previous one is not done. */
2785 seq = seq_read(connectivity_seq_get());
2786 if (seq != connectivity_seqno || status_txn_try_again) {
2789 connectivity_seqno = seq;
2790 status_txn = ovsdb_idl_txn_create(idl);
2791 HMAP_FOR_EACH (br, node, &all_bridges) {
2794 br_refresh_stp_status(br);
2795 br_refresh_rstp_status(br);
2796 br_refresh_datapath_info(br);
2797 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2798 struct iface *iface;
2800 port_refresh_stp_status(port);
2801 port_refresh_rstp_status(port);
2802 port_refresh_bond_status(port, status_txn_try_again);
2803 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2804 iface_refresh_netdev_status(iface);
2805 iface_refresh_ofproto_status(iface);
2812 /* Commit the transaction and get the status. If the transaction finishes,
2813 * then destroy the transaction. Otherwise, keep it so that we can check
2814 * progress the next time that this function is called. */
2816 enum ovsdb_idl_txn_status status;
2818 status = ovsdb_idl_txn_commit(status_txn);
2819 if (status != TXN_INCOMPLETE) {
2820 ovsdb_idl_txn_destroy(status_txn);
2823 /* Sets the 'status_txn_try_again' if the transaction fails. */
2824 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2825 status_txn_try_again = false;
2827 status_txn_try_again = true;
2832 /* Refresh AA port status if necessary. */
2833 if (time_msec() >= aa_refresh_timer) {
2836 HMAP_FOR_EACH (br, node, &all_bridges) {
2837 if (bridge_aa_need_refresh(br)) {
2838 struct ovsdb_idl_txn *txn;
2840 txn = ovsdb_idl_txn_create(idl);
2841 bridge_aa_refresh_queued(br);
2842 ovsdb_idl_txn_commit(txn);
2843 ovsdb_idl_txn_destroy(txn);
2847 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2852 status_update_wait(void)
2854 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2855 * transaction to complete. If the status update to database needs to be
2856 * run again (transaction fails), registers a timeout in
2857 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2858 * sequence number. */
2860 ovsdb_idl_txn_wait(status_txn);
2861 } else if (status_txn_try_again) {
2862 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2864 seq_wait(connectivity_seq_get(), connectivity_seqno);
2875 /* Let each datapath type do the work that it needs to do. */
2877 ofproto_enumerate_types(&types);
2878 SSET_FOR_EACH (type, &types) {
2879 ofproto_type_run(type);
2881 sset_destroy(&types);
2883 /* Let each bridge do the work that it needs to do. */
2884 HMAP_FOR_EACH (br, node, &all_bridges) {
2885 ofproto_run(br->ofproto);
2892 static struct ovsrec_open_vswitch null_cfg;
2893 const struct ovsrec_open_vswitch *cfg;
2895 bool vlan_splinters_changed;
2897 ovsrec_open_vswitch_init(&null_cfg);
2903 if (ovsdb_idl_is_lock_contended(idl)) {
2904 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2905 struct bridge *br, *next_br;
2907 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2908 "disabling this process (pid %ld) until it goes away",
2909 (long int) getpid());
2911 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2912 bridge_destroy(br, false);
2914 /* Since we will not be running system_stats_run() in this process
2915 * with the current situation of multiple ovs-vswitchd daemons,
2916 * disable system stats collection. */
2917 system_stats_enable(false);
2919 } else if (!ovsdb_idl_has_lock(idl)
2920 || !ovsdb_idl_has_ever_connected(idl)) {
2921 /* Returns if not holding the lock or not done retrieving db
2925 cfg = ovsrec_open_vswitch_first(idl);
2927 /* Initialize the ofproto library. This only needs to run once, but
2928 * it must be done after the configuration is set. If the
2929 * initialization has already occurred, bridge_init_ofproto()
2930 * returns immediately. */
2931 bridge_init_ofproto(cfg);
2933 /* Once the value of flow-restore-wait is false, we no longer should
2934 * check its value from the database. */
2935 if (cfg && ofproto_get_flow_restore_wait()) {
2936 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2937 "flow-restore-wait", false));
2942 /* Re-configure SSL. We do this on every trip through the main loop,
2943 * instead of just when the database changes, because the contents of the
2944 * key and certificate files can change without the database changing.
2946 * We do this before bridge_reconfigure() because that function might
2947 * initiate SSL connections and thus requires SSL to be configured. */
2948 if (cfg && cfg->ssl) {
2949 const struct ovsrec_ssl *ssl = cfg->ssl;
2951 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2952 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2955 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2956 * usage has changed. */
2957 vlan_splinters_changed = false;
2958 if (vlan_splinters_enabled_anywhere) {
2961 HMAP_FOR_EACH (br, node, &all_bridges) {
2962 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2963 vlan_splinters_changed = true;
2969 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed
2970 || ifaces_changed) {
2971 struct ovsdb_idl_txn *txn;
2973 ifaces_changed = false;
2975 idl_seqno = ovsdb_idl_get_seqno(idl);
2976 txn = ovsdb_idl_txn_create(idl);
2977 bridge_reconfigure(cfg ? cfg : &null_cfg);
2980 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2981 discover_types(cfg);
2984 /* If we are completing our initial configuration for this run
2985 * of ovs-vswitchd, then keep the transaction around to monitor
2986 * it for completion. */
2987 if (initial_config_done) {
2988 /* Always sets the 'status_txn_try_again' to check again,
2989 * in case that this transaction fails. */
2990 status_txn_try_again = true;
2991 ovsdb_idl_txn_commit(txn);
2992 ovsdb_idl_txn_destroy(txn);
2994 initial_config_done = true;
2995 daemonize_txn = txn;
2999 if (daemonize_txn) {
3000 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
3001 if (status != TXN_INCOMPLETE) {
3002 ovsdb_idl_txn_destroy(daemonize_txn);
3003 daemonize_txn = NULL;
3005 /* ovs-vswitchd has completed initialization, so allow the
3006 * process that forked us to exit successfully. */
3007 daemonize_complete();
3009 vlog_enable_async();
3011 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3016 run_status_update();
3026 ovsdb_idl_wait(idl);
3027 if (daemonize_txn) {
3028 ovsdb_idl_txn_wait(daemonize_txn);
3032 if (ifaces_changed) {
3033 poll_immediate_wake();
3037 ofproto_enumerate_types(&types);
3038 SSET_FOR_EACH (type, &types) {
3039 ofproto_type_wait(type);
3041 sset_destroy(&types);
3043 if (!hmap_is_empty(&all_bridges)) {
3046 HMAP_FOR_EACH (br, node, &all_bridges) {
3047 ofproto_wait(br->ofproto);
3049 stats_update_wait();
3050 status_update_wait();
3053 system_stats_wait();
3056 /* Adds some memory usage statistics for bridges into 'usage', for use with
3057 * memory_report(). */
3059 bridge_get_memory_usage(struct simap *usage)
3066 ofproto_enumerate_types(&types);
3067 SSET_FOR_EACH (type, &types) {
3068 ofproto_type_get_memory_usage(type, usage);
3070 sset_destroy(&types);
3072 HMAP_FOR_EACH (br, node, &all_bridges) {
3073 ofproto_get_memory_usage(br->ofproto, usage);
3077 /* QoS unixctl user interface functions. */
3079 struct qos_unixctl_show_cbdata {
3081 struct iface *iface;
3085 qos_unixctl_show_queue(unsigned int queue_id,
3086 const struct smap *details,
3087 struct iface *iface,
3090 struct netdev_queue_stats stats;
3091 struct smap_node *node;
3094 ds_put_cstr(ds, "\n");
3096 ds_put_format(ds, "Queue %u:\n", queue_id);
3098 ds_put_cstr(ds, "Default:\n");
3101 SMAP_FOR_EACH (node, details) {
3102 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3105 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3107 if (stats.tx_packets != UINT64_MAX) {
3108 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3111 if (stats.tx_bytes != UINT64_MAX) {
3112 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3115 if (stats.tx_errors != UINT64_MAX) {
3116 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3119 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3120 queue_id, ovs_strerror(error));
3125 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3126 const char *argv[], void *aux OVS_UNUSED)
3128 struct ds ds = DS_EMPTY_INITIALIZER;
3129 struct smap smap = SMAP_INITIALIZER(&smap);
3130 struct iface *iface;
3132 struct smap_node *node;
3134 iface = iface_find(argv[1]);
3136 unixctl_command_reply_error(conn, "no such interface");
3140 netdev_get_qos(iface->netdev, &type, &smap);
3142 if (*type != '\0') {
3143 struct netdev_queue_dump dump;
3144 struct smap details;
3145 unsigned int queue_id;
3147 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3149 SMAP_FOR_EACH (node, &smap) {
3150 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3153 smap_init(&details);
3154 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3155 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3157 smap_destroy(&details);
3159 unixctl_command_reply(conn, ds_cstr(&ds));
3161 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3162 unixctl_command_reply_error(conn, ds_cstr(&ds));
3165 smap_destroy(&smap);
3169 /* Bridge reconfiguration functions. */
3171 bridge_create(const struct ovsrec_bridge *br_cfg)
3175 ovs_assert(!bridge_lookup(br_cfg->name));
3176 br = xzalloc(sizeof *br);
3178 br->name = xstrdup(br_cfg->name);
3179 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3182 /* Derive the default Ethernet address from the bridge's UUID. This should
3183 * be unique and it will be stable between ovs-vswitchd runs. */
3184 memcpy(&br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3185 eth_addr_mark_random(&br->default_ea);
3187 hmap_init(&br->ports);
3188 hmap_init(&br->ifaces);
3189 hmap_init(&br->iface_by_name);
3190 hmap_init(&br->mirrors);
3192 hmap_init(&br->mappings);
3193 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3197 bridge_destroy(struct bridge *br, bool del)
3200 struct mirror *mirror, *next_mirror;
3201 struct port *port, *next_port;
3203 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3206 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3207 mirror_destroy(mirror);
3210 hmap_remove(&all_bridges, &br->node);
3211 ofproto_destroy(br->ofproto, del);
3212 hmap_destroy(&br->ifaces);
3213 hmap_destroy(&br->ports);
3214 hmap_destroy(&br->iface_by_name);
3215 hmap_destroy(&br->mirrors);
3216 hmap_destroy(&br->mappings);
3223 static struct bridge *
3224 bridge_lookup(const char *name)
3228 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3229 if (!strcmp(br->name, name)) {
3236 /* Handle requests for a listing of all flows known by the OpenFlow
3237 * stack, including those normally hidden. */
3239 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3240 const char *argv[], void *aux OVS_UNUSED)
3245 br = bridge_lookup(argv[1]);
3247 unixctl_command_reply_error(conn, "Unknown bridge");
3252 ofproto_get_all_flows(br->ofproto, &results);
3254 unixctl_command_reply(conn, ds_cstr(&results));
3255 ds_destroy(&results);
3258 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3259 * connections and reconnect. If BRIDGE is not specified, then all bridges
3260 * drop their controller connections and reconnect. */
3262 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3263 const char *argv[], void *aux OVS_UNUSED)
3267 br = bridge_lookup(argv[1]);
3269 unixctl_command_reply_error(conn, "Unknown bridge");
3272 ofproto_reconnect_controllers(br->ofproto);
3274 HMAP_FOR_EACH (br, node, &all_bridges) {
3275 ofproto_reconnect_controllers(br->ofproto);
3278 unixctl_command_reply(conn, NULL);
3282 bridge_get_controllers(const struct bridge *br,
3283 struct ovsrec_controller ***controllersp)
3285 struct ovsrec_controller **controllers;
3286 size_t n_controllers;
3288 controllers = br->cfg->controller;
3289 n_controllers = br->cfg->n_controller;
3291 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3297 *controllersp = controllers;
3299 return n_controllers;
3303 bridge_collect_wanted_ports(struct bridge *br,
3304 const unsigned long int *splinter_vlans,
3305 struct shash *wanted_ports)
3309 shash_init(wanted_ports);
3311 for (i = 0; i < br->cfg->n_ports; i++) {
3312 const char *name = br->cfg->ports[i]->name;
3313 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3314 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3318 if (bridge_get_controllers(br, NULL)
3319 && !shash_find(wanted_ports, br->name)) {
3320 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3321 br->name, br->name);
3323 ovsrec_interface_init(&br->synth_local_iface);
3324 ovsrec_port_init(&br->synth_local_port);
3326 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3327 br->synth_local_port.n_interfaces = 1;
3328 br->synth_local_port.name = br->name;
3330 br->synth_local_iface.name = br->name;
3331 br->synth_local_iface.type = "internal";
3333 br->synth_local_ifacep = &br->synth_local_iface;
3335 shash_add(wanted_ports, br->name, &br->synth_local_port);
3338 if (splinter_vlans) {
3339 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3343 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3344 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3345 * 'br' needs to complete its configuration. */
3347 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3349 struct shash_node *port_node;
3350 struct port *port, *next;
3352 /* Get rid of deleted ports.
3353 * Get rid of deleted interfaces on ports that still exist. */
3354 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3355 port->cfg = shash_find_data(wanted_ports, port->name);
3359 port_del_ifaces(port);
3363 /* Update iface->cfg and iface->type in interfaces that still exist. */
3364 SHASH_FOR_EACH (port_node, wanted_ports) {
3365 const struct ovsrec_port *port = port_node->data;
3368 for (i = 0; i < port->n_interfaces; i++) {
3369 const struct ovsrec_interface *cfg = port->interfaces[i];
3370 struct iface *iface = iface_lookup(br, cfg->name);
3371 const char *type = iface_get_type(cfg, br->cfg);
3376 } else if (!strcmp(type, "null")) {
3377 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3378 " may be removed in February 2013. Please email"
3379 " dev@openvswitch.org with concerns.",
3382 /* We will add new interfaces later. */
3388 /* Initializes 'oc' appropriately as a management service controller for
3391 * The caller must free oc->target when it is no longer needed. */
3393 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3394 struct ofproto_controller *oc)
3396 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3397 oc->max_backoff = 0;
3398 oc->probe_interval = 60;
3399 oc->band = OFPROTO_OUT_OF_BAND;
3401 oc->burst_limit = 0;
3402 oc->enable_async_msgs = true;
3406 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3408 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3409 struct ofproto_controller *oc)
3413 oc->target = c->target;
3414 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3415 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3416 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3417 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3418 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3419 oc->burst_limit = (c->controller_burst_limit
3420 ? *c->controller_burst_limit : 0);
3421 oc->enable_async_msgs = (!c->enable_async_messages
3422 || *c->enable_async_messages);
3423 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3424 if (dscp < 0 || dscp > 63) {
3425 dscp = DSCP_DEFAULT;
3430 /* Configures the IP stack for 'br''s local interface properly according to the
3431 * configuration in 'c'. */
3433 bridge_configure_local_iface_netdev(struct bridge *br,
3434 struct ovsrec_controller *c)
3436 struct netdev *netdev;
3437 struct in_addr mask, gateway;
3439 struct iface *local_iface;
3442 /* If there's no local interface or no IP address, give up. */
3443 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3444 if (!local_iface || !c->local_ip || !ip_parse(c->local_ip, &ip.s_addr)) {
3448 /* Bring up the local interface. */
3449 netdev = local_iface->netdev;
3450 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3452 /* Configure the IP address and netmask. */
3453 if (!c->local_netmask
3454 || !ip_parse(c->local_netmask, &mask.s_addr)
3456 mask.s_addr = guess_netmask(ip.s_addr);
3458 if (!netdev_set_in4(netdev, ip, mask)) {
3459 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3460 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3463 /* Configure the default gateway. */
3464 if (c->local_gateway
3465 && ip_parse(c->local_gateway, &gateway.s_addr)
3466 && gateway.s_addr) {
3467 if (!netdev_add_router(netdev, gateway)) {
3468 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3469 br->name, IP_ARGS(gateway.s_addr));
3474 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3475 * in either string are treated as equal to any number of slashes in the other,
3476 * e.g. "x///y" is equal to "x/y".
3478 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3479 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3480 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3481 * 'b' against a prefix of 'a'.
3484 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3486 const char *b_start = b;
3488 if (b - b_start >= b_stoplen) {
3490 } else if (*a != *b) {
3492 } else if (*a == '/') {
3493 a += strspn(a, "/");
3494 b += strspn(b, "/");
3495 } else if (*a == '\0') {
3505 bridge_configure_remotes(struct bridge *br,
3506 const struct sockaddr_in *managers, size_t n_managers)
3508 bool disable_in_band;
3510 struct ovsrec_controller **controllers;
3511 size_t n_controllers;
3513 enum ofproto_fail_mode fail_mode;
3515 struct ofproto_controller *ocs;
3519 /* Check if we should disable in-band control on this bridge. */
3520 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3523 /* Set OpenFlow queue ID for in-band control. */
3524 ofproto_set_in_band_queue(br->ofproto,
3525 smap_get_int(&br->cfg->other_config,
3526 "in-band-queue", -1));
3528 if (disable_in_band) {
3529 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3531 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3534 n_controllers = bridge_get_controllers(br, &controllers);
3536 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3539 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3540 for (i = 0; i < n_controllers; i++) {
3541 struct ovsrec_controller *c = controllers[i];
3543 if (!strncmp(c->target, "punix:", 6)
3544 || !strncmp(c->target, "unix:", 5)) {
3545 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3548 if (!strncmp(c->target, "unix:", 5)) {
3549 /* Connect to a listening socket */
3550 whitelist = xasprintf("unix:%s/", ovs_rundir());
3551 if (strchr(c->target, '/') &&
3552 !equal_pathnames(c->target, whitelist,
3553 strlen(whitelist))) {
3554 /* Absolute path specified, but not in ovs_rundir */
3555 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3556 "controller \"%s\" due to possibility for "
3557 "remote exploit. Instead, specify socket "
3558 "in whitelisted \"%s\" or connect to "
3559 "\"unix:%s/%s.mgmt\" (which is always "
3560 "available without special configuration).",
3561 br->name, c->target, whitelist,
3562 ovs_rundir(), br->name);
3567 whitelist = xasprintf("punix:%s/%s.",
3568 ovs_rundir(), br->name);
3569 if (!equal_pathnames(c->target, whitelist, strlen(whitelist))
3570 || strchr(c->target + strlen(whitelist), '/')) {
3571 /* Prevent remote ovsdb-server users from accessing
3572 * arbitrary Unix domain sockets and overwriting arbitrary
3574 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3575 "controller \"%s\" due to possibility of "
3576 "overwriting local files. Instead, specify "
3577 "path in whitelisted format \"%s*\" or "
3578 "connect to \"unix:%s/%s.mgmt\" (which is "
3579 "always available without special "
3581 br->name, c->target, whitelist,
3582 ovs_rundir(), br->name);
3591 bridge_configure_local_iface_netdev(br, c);
3592 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3593 if (disable_in_band) {
3594 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3599 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3600 bridge_get_allowed_versions(br));
3601 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3604 /* Set the fail-mode. */
3605 fail_mode = !br->cfg->fail_mode
3606 || !strcmp(br->cfg->fail_mode, "standalone")
3607 ? OFPROTO_FAIL_STANDALONE
3608 : OFPROTO_FAIL_SECURE;
3609 ofproto_set_fail_mode(br->ofproto, fail_mode);
3611 /* Configure OpenFlow controller connection snooping. */
3612 if (!ofproto_has_snoops(br->ofproto)) {
3616 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3617 ovs_rundir(), br->name));
3618 ofproto_set_snoops(br->ofproto, &snoops);
3619 sset_destroy(&snoops);
3624 bridge_configure_tables(struct bridge *br)
3626 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3630 n_tables = ofproto_get_n_tables(br->ofproto);
3632 for (i = 0; i < n_tables; i++) {
3633 struct ofproto_table_settings s;
3634 bool use_default_prefixes = true;
3637 s.max_flows = UINT_MAX;
3639 s.enable_eviction = false;
3641 s.n_prefix_fields = 0;
3642 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3644 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3645 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3648 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3649 s.max_flows = *cfg->flow_limit;
3652 s.enable_eviction = (cfg->overflow_policy
3653 && !strcmp(cfg->overflow_policy, "evict"));
3654 if (cfg->n_groups) {
3655 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3656 for (k = 0; k < cfg->n_groups; k++) {
3657 const char *string = cfg->groups[k];
3660 msg = mf_parse_subfield__(&s.groups[k], &string);
3662 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3663 "'groups' (%s)", br->name, i, msg);
3665 } else if (*string) {
3666 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3667 "element '%s' contains trailing garbage",
3668 br->name, i, cfg->groups[k]);
3675 /* Prefix lookup fields. */
3676 s.n_prefix_fields = 0;
3677 for (k = 0; k < cfg->n_prefixes; k++) {
3678 const char *name = cfg->prefixes[k];
3679 const struct mf_field *mf;
3681 if (strcmp(name, "none") == 0) {
3682 use_default_prefixes = false;
3683 s.n_prefix_fields = 0;
3686 mf = mf_from_name(name);
3688 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3692 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3693 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3694 "%s", br->name, name);
3697 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3698 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3699 "field not used: %s", br->name, name);
3702 use_default_prefixes = false;
3703 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3706 if (use_default_prefixes) {
3707 /* Use default values. */
3708 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3709 memcpy(s.prefix_fields, default_prefix_fields,
3710 sizeof default_prefix_fields);
3713 struct ds ds = DS_EMPTY_INITIALIZER;
3714 for (k = 0; k < s.n_prefix_fields; k++) {
3716 ds_put_char(&ds, ',');
3718 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3720 if (s.n_prefix_fields == 0) {
3721 ds_put_cstr(&ds, "none");
3723 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3724 br->name, i, ds_cstr(&ds));
3728 ofproto_configure_table(br->ofproto, i, &s);
3732 for (; j < br->cfg->n_flow_tables; j++) {
3733 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3734 "%"PRId64" not supported by this datapath", br->name,
3735 br->cfg->key_flow_tables[j]);
3740 bridge_configure_dp_desc(struct bridge *br)
3742 ofproto_set_dp_desc(br->ofproto,
3743 smap_get(&br->cfg->other_config, "dp-desc"));
3746 static struct aa_mapping *
3747 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3749 struct aa_mapping *m;
3751 HMAP_FOR_EACH_IN_BUCKET (m,
3753 hash_bytes(&isid, sizeof isid, 0),
3755 if (isid == m->isid) {
3762 static struct aa_mapping *
3763 bridge_aa_mapping_create(struct bridge *br,
3767 struct aa_mapping *m;
3769 m = xzalloc(sizeof *m);
3773 m->br_name = xstrdup(br->name);
3774 hmap_insert(&br->mappings,
3776 hash_bytes(&isid, sizeof isid, 0));
3782 bridge_aa_mapping_destroy(struct aa_mapping *m)
3785 struct bridge *br = m->bridge;
3788 ofproto_aa_mapping_unregister(br->ofproto, m);
3791 hmap_remove(&br->mappings, &m->hmap_node);
3800 bridge_aa_mapping_configure(struct aa_mapping *m)
3802 struct aa_mapping_settings s;
3808 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3814 bridge_configure_aa(struct bridge *br)
3816 const struct ovsdb_datum *mc;
3817 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3818 struct aa_settings aa_s;
3819 struct aa_mapping *m, *next;
3823 ofproto_set_aa(br->ofproto, NULL, NULL);
3827 memset(&aa_s, 0, sizeof aa_s);
3828 aa_s.system_description = auto_attach->system_description;
3829 aa_s.system_name = auto_attach->system_name;
3830 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3832 mc = ovsrec_autoattach_get_mappings(auto_attach,
3834 OVSDB_TYPE_INTEGER);
3835 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3836 union ovsdb_atom atom;
3838 atom.integer = m->isid;
3839 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3840 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3842 bridge_aa_mapping_destroy(m);
3846 /* Add new mappings and reconfigure existing ones. */
3847 for (i = 0; i < auto_attach->n_mappings; ++i) {
3848 struct aa_mapping *m =
3849 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3852 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3853 auto_attach->key_mappings[i],
3854 auto_attach->value_mappings[i]);
3855 m = bridge_aa_mapping_create(br,
3856 auto_attach->key_mappings[i],
3857 auto_attach->value_mappings[i]);
3859 if (!bridge_aa_mapping_configure(m)) {
3860 bridge_aa_mapping_destroy(m);
3867 bridge_aa_need_refresh(struct bridge *br)
3869 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3873 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3875 int64_t *trunks = NULL;
3877 bool found = false, reconfigure = false;
3879 for (i = 0; i < port->cfg->n_trunks; i++) {
3880 if (port->cfg->trunks[i] == m->vlan) {
3887 case BRIDGE_AA_VLAN_OPER_ADD:
3889 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3891 for (i = 0; i < port->cfg->n_trunks; i++) {
3892 trunks[i] = port->cfg->trunks[i];
3894 trunks[i++] = m->vlan;
3900 case BRIDGE_AA_VLAN_OPER_REMOVE:
3904 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3906 for (i = 0; i < port->cfg->n_trunks; i++) {
3907 if (port->cfg->trunks[i] != m->vlan) {
3908 trunks[j++] = port->cfg->trunks[i];
3917 case BRIDGE_AA_VLAN_OPER_UNDEF:
3919 VLOG_WARN("unrecognized operation %u", m->oper);
3924 /* VLAN switching under trunk mode cause the trunk port to switch all
3925 * VLANs, see ovs-vswitchd.conf.db
3928 static char *vlan_mode_access = "access";
3929 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3933 static char *vlan_mode_trunk = "trunk";
3934 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3937 ovsrec_port_set_trunks(port->cfg, trunks, i);
3939 /* Force reconfigure of the port. */
3940 port_configure(port);
3945 bridge_aa_refresh_queued(struct bridge *br)
3947 struct ovs_list *list = xmalloc(sizeof *list);
3948 struct bridge_aa_vlan *node, *next;
3951 ofproto_aa_vlan_get_queued(br->ofproto, list);
3953 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3956 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3959 port = port_lookup(br, node->port_name);
3961 bridge_aa_update_trunks(port, node);
3964 list_remove(&node->list_node);
3965 free(node->port_name);
3973 /* Port functions. */
3975 static struct port *
3976 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3980 port = xzalloc(sizeof *port);
3982 port->name = xstrdup(cfg->name);
3984 list_init(&port->ifaces);
3986 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3990 /* Deletes interfaces from 'port' that are no longer configured for it. */
3992 port_del_ifaces(struct port *port)
3994 struct iface *iface, *next;
3995 struct sset new_ifaces;
3998 /* Collect list of new interfaces. */
3999 sset_init(&new_ifaces);
4000 for (i = 0; i < port->cfg->n_interfaces; i++) {
4001 const char *name = port->cfg->interfaces[i]->name;
4002 const char *type = port->cfg->interfaces[i]->type;
4003 if (strcmp(type, "null")) {
4004 sset_add(&new_ifaces, name);
4008 /* Get rid of deleted interfaces. */
4009 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4010 if (!sset_contains(&new_ifaces, iface->name)) {
4011 iface_destroy(iface);
4015 sset_destroy(&new_ifaces);
4019 port_destroy(struct port *port)
4022 struct bridge *br = port->bridge;
4023 struct iface *iface, *next;
4026 ofproto_bundle_unregister(br->ofproto, port);
4029 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4030 iface_destroy__(iface);
4033 hmap_remove(&br->ports, &port->hmap_node);
4039 static struct port *
4040 port_lookup(const struct bridge *br, const char *name)
4044 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4046 if (!strcmp(port->name, name)) {
4054 enable_lacp(struct port *port, bool *activep)
4056 if (!port->cfg->lacp) {
4057 /* XXX when LACP implementation has been sufficiently tested, enable by
4058 * default and make active on bonded ports. */
4060 } else if (!strcmp(port->cfg->lacp, "off")) {
4062 } else if (!strcmp(port->cfg->lacp, "active")) {
4065 } else if (!strcmp(port->cfg->lacp, "passive")) {
4069 VLOG_WARN("port %s: unknown LACP mode %s",
4070 port->name, port->cfg->lacp);
4075 static struct lacp_settings *
4076 port_configure_lacp(struct port *port, struct lacp_settings *s)
4078 const char *lacp_time, *system_id;
4081 if (!enable_lacp(port, &s->active)) {
4085 s->name = port->name;
4087 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4089 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4090 ETH_ADDR_SCAN_ARGS(s->id))) {
4091 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4092 " address.", port->name, system_id);
4096 s->id = port->bridge->ea;
4099 if (eth_addr_is_zero(s->id)) {
4100 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4104 /* Prefer bondable links if unspecified. */
4105 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4107 s->priority = (priority > 0 && priority <= UINT16_MAX
4109 : UINT16_MAX - !list_is_short(&port->ifaces));
4111 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4112 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4114 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4115 "lacp-fallback-ab", false);
4121 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4123 int priority, portid, key;
4125 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4126 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4128 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4130 if (portid <= 0 || portid > UINT16_MAX) {
4131 portid = ofp_to_u16(iface->ofp_port);
4134 if (priority <= 0 || priority > UINT16_MAX) {
4135 priority = UINT16_MAX;
4138 if (key < 0 || key > UINT16_MAX) {
4142 s->name = iface->name;
4144 s->priority = priority;
4149 port_configure_bond(struct port *port, struct bond_settings *s)
4151 const char *detect_s;
4152 struct iface *iface;
4154 int miimon_interval;
4156 s->name = port->name;
4158 if (port->cfg->bond_mode) {
4159 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4160 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4161 port->name, port->cfg->bond_mode,
4162 bond_mode_to_string(s->balance));
4165 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4167 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4168 * active-backup. At some point we should remove this warning. */
4169 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4170 " in previous versions, the default bond_mode was"
4171 " balance-slb", port->name,
4172 bond_mode_to_string(s->balance));
4174 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4175 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4176 "please use another bond type or disable flood_vlans",
4180 miimon_interval = smap_get_int(&port->cfg->other_config,
4181 "bond-miimon-interval", 0);
4182 if (miimon_interval <= 0) {
4183 miimon_interval = 200;
4186 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4187 if (!detect_s || !strcmp(detect_s, "carrier")) {
4188 miimon_interval = 0;
4189 } else if (strcmp(detect_s, "miimon")) {
4190 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4191 "defaulting to carrier", port->name, detect_s);
4192 miimon_interval = 0;
4195 s->up_delay = MAX(0, port->cfg->bond_updelay);
4196 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4197 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4198 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4199 "bond-rebalance-interval", 10000);
4200 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4201 s->rebalance_interval = 1000;
4204 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4205 "lacp-fallback-ab", false);
4207 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4208 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4211 mac_s = port->cfg->bond_active_slave;
4212 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4213 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4214 /* OVSDB did not store the last active interface */
4215 s->active_slave_mac = eth_addr_zero;
4219 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4220 * instead of obtaining it from the database. */
4222 port_is_synthetic(const struct port *port)
4224 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4227 /* Interface functions. */
4230 iface_is_internal(const struct ovsrec_interface *iface,
4231 const struct ovsrec_bridge *br)
4233 /* The local port and "internal" ports are always "internal". */
4234 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4237 /* Returns the correct network device type for interface 'iface' in bridge
4240 iface_get_type(const struct ovsrec_interface *iface,
4241 const struct ovsrec_bridge *br)
4245 /* The local port always has type "internal". Other ports take
4246 * their type from the database and default to "system" if none is
4248 if (iface_is_internal(iface, br)) {
4251 type = iface->type[0] ? iface->type : "system";
4254 return ofproto_port_open_type(br->datapath_type, type);
4258 iface_destroy__(struct iface *iface)
4261 struct port *port = iface->port;
4262 struct bridge *br = port->bridge;
4264 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4265 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4268 if (iface->ofp_port != OFPP_NONE) {
4269 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4272 list_remove(&iface->port_elem);
4273 hmap_remove(&br->iface_by_name, &iface->name_node);
4275 /* The user is changing configuration here, so netdev_remove needs to be
4276 * used as opposed to netdev_close */
4277 netdev_remove(iface->netdev);
4285 iface_destroy(struct iface *iface)
4288 struct port *port = iface->port;
4290 iface_destroy__(iface);
4291 if (list_is_empty(&port->ifaces)) {
4297 static struct iface *
4298 iface_lookup(const struct bridge *br, const char *name)
4300 struct iface *iface;
4302 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4303 &br->iface_by_name) {
4304 if (!strcmp(iface->name, name)) {
4312 static struct iface *
4313 iface_find(const char *name)
4315 const struct bridge *br;
4317 HMAP_FOR_EACH (br, node, &all_bridges) {
4318 struct iface *iface = iface_lookup(br, name);
4327 static struct iface *
4328 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4330 struct iface *iface;
4332 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4334 if (iface->ofp_port == ofp_port) {
4341 /* Set Ethernet address of 'iface', if one is specified in the configuration
4344 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4346 struct eth_addr ea, *mac = NULL;
4347 struct iface *hw_addr_iface;
4349 if (strcmp(iface->type, "internal")) {
4353 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, &ea)) {
4355 } else if (port->cfg->fake_bridge) {
4356 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4357 find_local_hw_addr(br, &ea, port, &hw_addr_iface);
4362 if (iface->ofp_port == OFPP_LOCAL) {
4363 VLOG_ERR("interface %s: ignoring mac in Interface record "
4364 "(use Bridge record to set local port's mac)",
4366 } else if (eth_addr_is_multicast(*mac)) {
4367 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4370 int error = netdev_set_etheraddr(iface->netdev, *mac);
4372 VLOG_ERR("interface %s: setting MAC failed (%s)",
4373 iface->name, ovs_strerror(error));
4379 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4381 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4383 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4384 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4385 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4389 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4390 * sets the "ofport" field to -1.
4392 * This is appropriate when 'if_cfg''s interface cannot be created or is
4393 * otherwise invalid. */
4395 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4397 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4398 iface_set_ofport(if_cfg, OFPP_NONE);
4399 ovsrec_interface_set_error(if_cfg, errp);
4400 ovsrec_interface_set_status(if_cfg, NULL);
4401 ovsrec_interface_set_admin_state(if_cfg, NULL);
4402 ovsrec_interface_set_duplex(if_cfg, NULL);
4403 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4404 ovsrec_interface_set_link_state(if_cfg, NULL);
4405 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4406 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4407 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4408 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4409 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4410 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4411 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4412 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4417 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4419 union ovsdb_atom atom;
4421 atom.integer = target;
4422 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4426 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4428 struct ofpbuf queues_buf;
4430 ofpbuf_init(&queues_buf, 0);
4432 if (!qos || qos->type[0] == '\0') {
4433 netdev_set_qos(iface->netdev, NULL, NULL);
4435 const struct ovsdb_datum *queues;
4436 struct netdev_queue_dump dump;
4437 unsigned int queue_id;
4438 struct smap details;
4442 /* Configure top-level Qos for 'iface'. */
4443 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4445 /* Deconfigure queues that were deleted. */
4446 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4448 smap_init(&details);
4449 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4450 if (!queue_ids_include(queues, queue_id)) {
4451 netdev_delete_queue(iface->netdev, queue_id);
4454 smap_destroy(&details);
4456 /* Configure queues for 'iface'. */
4458 for (i = 0; i < qos->n_queues; i++) {
4459 const struct ovsrec_queue *queue = qos->value_queues[i];
4460 unsigned int queue_id = qos->key_queues[i];
4462 if (queue_id == 0) {
4466 if (queue->n_dscp == 1) {
4467 struct ofproto_port_queue *port_queue;
4469 port_queue = ofpbuf_put_uninit(&queues_buf,
4470 sizeof *port_queue);
4471 port_queue->queue = queue_id;
4472 port_queue->dscp = queue->dscp[0];
4475 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4478 struct smap details;
4480 smap_init(&details);
4481 netdev_set_queue(iface->netdev, 0, &details);
4482 smap_destroy(&details);
4486 if (iface->ofp_port != OFPP_NONE) {
4487 const struct ofproto_port_queue *port_queues = queues_buf.data;
4488 size_t n_queues = queues_buf.size / sizeof *port_queues;
4490 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4491 port_queues, n_queues);
4494 netdev_set_policing(iface->netdev,
4495 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4496 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4498 ofpbuf_uninit(&queues_buf);
4502 iface_configure_cfm(struct iface *iface)
4504 const struct ovsrec_interface *cfg = iface->cfg;
4505 const char *opstate_str;
4506 const char *cfm_ccm_vlan;
4507 struct cfm_settings s;
4508 struct smap netdev_args;
4510 if (!cfg->n_cfm_mpid) {
4511 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4515 s.check_tnl_key = false;
4516 smap_init(&netdev_args);
4517 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4518 const char *key = smap_get(&netdev_args, "key");
4519 const char *in_key = smap_get(&netdev_args, "in_key");
4521 s.check_tnl_key = (key && !strcmp(key, "flow"))
4522 || (in_key && !strcmp(in_key, "flow"));
4524 smap_destroy(&netdev_args);
4526 s.mpid = *cfg->cfm_mpid;
4527 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4528 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4529 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4531 if (s.interval <= 0) {
4535 if (!cfm_ccm_vlan) {
4537 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4538 s.ccm_vlan = CFM_RANDOM_VLAN;
4540 s.ccm_vlan = atoi(cfm_ccm_vlan);
4541 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4546 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4548 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4550 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4551 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4553 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4556 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4557 * instead of obtaining it from the database. */
4559 iface_is_synthetic(const struct iface *iface)
4561 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4565 iface_validate_ofport__(size_t n, int64_t *ofport)
4567 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4568 ? u16_to_ofp(*ofport)
4573 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4575 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4579 iface_pick_ofport(const struct ovsrec_interface *cfg)
4581 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4582 return (requested_ofport != OFPP_NONE
4584 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4587 /* Port mirroring. */
4589 static struct mirror *
4590 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4594 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4595 if (uuid_equals(uuid, &m->uuid)) {
4603 bridge_configure_mirrors(struct bridge *br)
4605 const struct ovsdb_datum *mc;
4606 unsigned long *flood_vlans;
4607 struct mirror *m, *next;
4610 /* Get rid of deleted mirrors. */
4611 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4612 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4613 union ovsdb_atom atom;
4615 atom.uuid = m->uuid;
4616 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4621 /* Add new mirrors and reconfigure existing ones. */
4622 for (i = 0; i < br->cfg->n_mirrors; i++) {
4623 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4624 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4626 m = mirror_create(br, cfg);
4629 if (!mirror_configure(m)) {
4634 /* Update flooded vlans (for RSPAN). */
4635 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4636 br->cfg->n_flood_vlans);
4637 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4638 bitmap_free(flood_vlans);
4641 static struct mirror *
4642 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4646 m = xzalloc(sizeof *m);
4647 m->uuid = cfg->header_.uuid;
4648 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4650 m->name = xstrdup(cfg->name);
4656 mirror_destroy(struct mirror *m)
4659 struct bridge *br = m->bridge;
4662 ofproto_mirror_unregister(br->ofproto, m);
4665 hmap_remove(&br->mirrors, &m->hmap_node);
4672 mirror_collect_ports(struct mirror *m,
4673 struct ovsrec_port **in_ports, int n_in_ports,
4674 void ***out_portsp, size_t *n_out_portsp)
4676 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4677 size_t n_out_ports = 0;
4680 for (i = 0; i < n_in_ports; i++) {
4681 const char *name = in_ports[i]->name;
4682 struct port *port = port_lookup(m->bridge, name);
4684 out_ports[n_out_ports++] = port;
4686 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4687 "port %s", m->bridge->name, m->name, name);
4690 *out_portsp = out_ports;
4691 *n_out_portsp = n_out_ports;
4695 mirror_configure(struct mirror *m)
4697 const struct ovsrec_mirror *cfg = m->cfg;
4698 struct ofproto_mirror_settings s;
4701 if (strcmp(cfg->name, m->name)) {
4703 m->name = xstrdup(cfg->name);
4707 /* Get output port or VLAN. */
4708 if (cfg->output_port) {
4709 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4710 if (!s.out_bundle) {
4711 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4712 m->bridge->name, m->name);
4715 s.out_vlan = UINT16_MAX;
4717 if (cfg->output_vlan) {
4718 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4719 "output vlan; ignoring output vlan",
4720 m->bridge->name, m->name);
4722 } else if (cfg->output_vlan) {
4723 /* The database should prevent invalid VLAN values. */
4724 s.out_bundle = NULL;
4725 s.out_vlan = *cfg->output_vlan;
4727 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4728 m->bridge->name, m->name);
4732 /* Get port selection. */
4733 if (cfg->select_all) {
4734 size_t n_ports = hmap_count(&m->bridge->ports);
4735 void **ports = xmalloc(n_ports * sizeof *ports);
4740 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4750 /* Get ports, dropping ports that don't exist.
4751 * The IDL ensures that there are no duplicates. */
4752 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4753 &s.srcs, &s.n_srcs);
4754 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4755 &s.dsts, &s.n_dsts);
4758 /* Get VLAN selection. */
4759 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4762 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4765 if (s.srcs != s.dsts) {
4774 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4776 * This is deprecated. It is only for compatibility with broken device drivers
4777 * in old versions of Linux that do not properly support VLANs when VLAN
4778 * devices are not used. When broken device drivers are no longer in
4779 * widespread use, we will delete these interfaces. */
4781 static struct ovsrec_port **recs;
4782 static size_t n_recs, allocated_recs;
4784 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4785 * splinters are reconfigured. */
4787 register_rec(struct ovsrec_port *rec)
4789 if (n_recs >= allocated_recs) {
4790 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4792 recs[n_recs++] = rec;
4795 /* Frees all of the ports registered with register_reg(). */
4797 free_registered_recs(void)
4801 for (i = 0; i < n_recs; i++) {
4802 struct ovsrec_port *port = recs[i];
4805 for (j = 0; j < port->n_interfaces; j++) {
4806 struct ovsrec_interface *iface = port->interfaces[j];
4811 smap_destroy(&port->other_config);
4812 free(port->interfaces);
4820 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4823 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4825 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4829 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4832 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4833 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4834 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4837 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4838 * use, returns NULL.
4840 * Updates 'vlan_splinters_enabled_anywhere'. */
4841 static unsigned long int *
4842 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4844 unsigned long int *splinter_vlans;
4845 struct sset splinter_ifaces;
4846 const char *real_dev_name;
4847 struct shash *real_devs;
4848 struct shash_node *node;
4852 /* Free space allocated for synthesized ports and interfaces, since we're
4853 * in the process of reconstructing all of them. */
4854 free_registered_recs();
4856 splinter_vlans = bitmap_allocate(4096);
4857 sset_init(&splinter_ifaces);
4858 vlan_splinters_enabled_anywhere = false;
4859 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4860 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4863 for (j = 0; j < br_cfg->n_ports; j++) {
4864 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4867 for (k = 0; k < port_cfg->n_interfaces; k++) {
4868 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4870 if (vlan_splinters_is_enabled(iface_cfg)) {
4871 vlan_splinters_enabled_anywhere = true;
4872 sset_add(&splinter_ifaces, iface_cfg->name);
4873 vlan_bitmap_from_array__(port_cfg->trunks,
4879 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4880 bitmap_set1(splinter_vlans, *port_cfg->tag);
4885 if (!vlan_splinters_enabled_anywhere) {
4886 free(splinter_vlans);
4887 sset_destroy(&splinter_ifaces);
4891 HMAP_FOR_EACH (br, node, &all_bridges) {
4893 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4897 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4898 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4899 * device to be created for it. */
4900 bitmap_set0(splinter_vlans, 0);
4901 bitmap_set0(splinter_vlans, 4095);
4903 /* Delete all VLAN devices that we don't need. */
4905 real_devs = vlandev_get_real_devs();
4906 SHASH_FOR_EACH (node, real_devs) {
4907 const struct vlan_real_dev *real_dev = node->data;
4908 const struct vlan_dev *vlan_dev;
4909 bool real_dev_has_splinters;
4911 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4913 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4914 if (!real_dev_has_splinters
4915 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4916 struct netdev *netdev;
4918 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4919 if (!netdev_get_in4(netdev, NULL, NULL) ||
4920 !netdev_get_in6(netdev, NULL)) {
4921 /* It has an IP address configured, so we don't own
4922 * it. Don't delete it. */
4924 vlandev_del(vlan_dev->name);
4926 netdev_close(netdev);
4933 /* Add all VLAN devices that we need. */
4934 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4937 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4938 if (!vlandev_get_name(real_dev_name, vid)) {
4939 vlandev_add(real_dev_name, vid);
4946 sset_destroy(&splinter_ifaces);
4948 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4949 free(splinter_vlans);
4952 return splinter_vlans;
4955 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4958 configure_splinter_port(struct port *port)
4960 struct ofproto *ofproto = port->bridge->ofproto;
4961 ofp_port_t realdev_ofp_port;
4962 const char *realdev_name;
4963 struct iface *vlandev, *realdev;
4965 ofproto_bundle_unregister(port->bridge->ofproto, port);
4967 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4970 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4971 realdev = iface_lookup(port->bridge, realdev_name);
4972 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4974 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4978 static struct ovsrec_port *
4979 synthesize_splinter_port(const char *real_dev_name,
4980 const char *vlan_dev_name, int vid)
4982 struct ovsrec_interface *iface;
4983 struct ovsrec_port *port;
4985 iface = xmalloc(sizeof *iface);
4986 ovsrec_interface_init(iface);
4987 iface->name = xstrdup(vlan_dev_name);
4988 iface->type = "system";
4990 port = xmalloc(sizeof *port);
4991 ovsrec_port_init(port);
4992 port->interfaces = xmemdup(&iface, sizeof iface);
4993 port->n_interfaces = 1;
4994 port->name = xstrdup(vlan_dev_name);
4995 port->vlan_mode = "splinter";
4996 port->tag = xmalloc(sizeof *port->tag);
4999 smap_add(&port->other_config, "realdev", real_dev_name);
5005 /* For each interface with 'br' that has VLAN splinters enabled, adds a
5006 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
5007 * 1-bit in the 'splinter_vlans' bitmap. */
5009 add_vlan_splinter_ports(struct bridge *br,
5010 const unsigned long int *splinter_vlans,
5011 struct shash *ports)
5015 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
5016 * we're modifying 'ports'. */
5017 for (i = 0; i < br->cfg->n_ports; i++) {
5018 const char *name = br->cfg->ports[i]->name;
5019 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
5022 for (j = 0; j < port_cfg->n_interfaces; j++) {
5023 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
5025 if (vlan_splinters_is_enabled(iface_cfg)) {
5026 const char *real_dev_name;
5029 real_dev_name = iface_cfg->name;
5030 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
5031 const char *vlan_dev_name;
5033 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5035 && !shash_find(ports, vlan_dev_name)) {
5036 shash_add(ports, vlan_dev_name,
5037 synthesize_splinter_port(
5038 real_dev_name, vlan_dev_name, vid));
5047 mirror_refresh_stats(struct mirror *m)
5049 struct ofproto *ofproto = m->bridge->ofproto;
5050 uint64_t tx_packets, tx_bytes;
5051 const char *keys[2];
5053 size_t stat_cnt = 0;
5055 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5056 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5060 if (tx_packets != UINT64_MAX) {
5061 keys[stat_cnt] = "tx_packets";
5062 values[stat_cnt] = tx_packets;
5065 if (tx_bytes != UINT64_MAX) {
5066 keys[stat_cnt] = "tx_bytes";
5067 values[stat_cnt] = tx_bytes;
5071 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5075 * Add registered netdev and dpif types to ovsdb to allow external
5076 * applications to query the capabilities of the Open vSwitch instance
5077 * running on the node.
5080 discover_types(const struct ovsrec_open_vswitch *cfg)
5084 /* Datapath types. */
5086 dp_enumerate_types(&types);
5087 const char **datapath_types = sset_array(&types);
5088 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5089 sset_count(&types));
5090 free(datapath_types);
5091 sset_destroy(&types);
5095 netdev_enumerate_types(&types);
5096 const char **iface_types = sset_array(&types);
5097 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5099 sset_destroy(&types);