1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
51 #include "if-notifier.h"
56 #include "socket-util.h"
58 #include "stream-ssl.h"
60 #include "system-stats.h"
65 #include "lib/vswitch-idl.h"
66 #include "xenserver.h"
67 #include "openvswitch/vlog.h"
68 #include "sflow_api.h"
69 #include "vlan-bitmap.h"
72 VLOG_DEFINE_THIS_MODULE(bridge);
74 COVERAGE_DEFINE(bridge_reconfigure);
77 /* These members are always valid.
79 * They are immutable: they never change between iface_create() and
81 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
86 struct netdev *netdev; /* Network device. */
87 ofp_port_t ofp_port; /* OpenFlow port number. */
90 /* These members are valid only within bridge_reconfigure(). */
91 const char *type; /* Usually same as cfg->type. */
92 const struct ovsrec_interface *cfg;
96 struct uuid uuid; /* UUID of this "mirror" record in database. */
97 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
98 struct bridge *bridge;
100 const struct ovsrec_mirror *cfg;
104 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
105 struct bridge *bridge;
108 const struct ovsrec_port *cfg;
110 /* An ordinary bridge port has 1 interface.
111 * A bridge port for bonding has at least 2 interfaces. */
112 struct ovs_list ifaces; /* List of "struct iface"s. */
116 struct hmap_node node; /* In 'all_bridges'. */
117 char *name; /* User-specified arbitrary name. */
118 char *type; /* Datapath type. */
119 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
120 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
121 const struct ovsrec_bridge *cfg;
123 /* OpenFlow switch processing. */
124 struct ofproto *ofproto; /* OpenFlow switch. */
127 struct hmap ports; /* "struct port"s indexed by name. */
128 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
129 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
131 /* Port mirroring. */
132 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
135 struct hmap mappings; /* "struct" indexed by UUID */
137 /* Used during reconfiguration. */
138 struct shash wanted_ports;
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
147 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
148 struct bridge *bridge;
154 /* All bridges, indexed by name. */
155 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
157 /* OVSDB IDL used to obtain configuration. */
158 static struct ovsdb_idl *idl;
160 /* We want to complete daemonization, fully detaching from our parent process,
161 * only after we have completed our initial configuration, committed our state
162 * to the database, and received confirmation back from the database server
163 * that it applied the commit. This allows our parent process to know that,
164 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
165 * to have already been filled in. (It is only "very likely" rather than
166 * certain because there is always a slim possibility that the transaction will
167 * fail or that some other client has added new bridges, ports, etc. while
168 * ovs-vswitchd was configuring using an old configuration.)
170 * We only need to do this once for our initial configuration at startup, so
171 * 'initial_config_done' tracks whether we've already done it. While we are
172 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
173 * itself and is otherwise NULL. */
174 static bool initial_config_done;
175 static struct ovsdb_idl_txn *daemonize_txn;
177 /* Most recently processed IDL sequence number. */
178 static unsigned int idl_seqno;
180 /* Track changes to port connectivity. */
181 static uint64_t connectivity_seqno = LLONG_MIN;
183 /* Status update to database.
185 * Some information in the database must be kept as up-to-date as possible to
186 * allow controllers to respond rapidly to network outages. Those status are
187 * updated via the 'status_txn'.
189 * We use the global connectivity sequence number to detect the status change.
190 * Also, to prevent the status update from sending too much to the database,
191 * we check the return status of each update transaction and do not start new
192 * update if the previous transaction status is 'TXN_INCOMPLETE'.
194 * 'statux_txn' is NULL if there is no ongoing status update.
196 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
197 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
198 * which will cause the main thread wake up soon and retry the status update.
200 static struct ovsdb_idl_txn *status_txn;
201 static bool status_txn_try_again;
203 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
204 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
205 #define STATUS_CHECK_AGAIN_MSEC 100
207 /* Statistics update to database. */
208 static struct ovsdb_idl_txn *stats_txn;
210 /* Each time this timer expires, the bridge fetches interface and mirror
211 * statistics and pushes them into the database. */
212 static int stats_timer_interval;
213 static long long int stats_timer = LLONG_MIN;
215 /* Each time this timer expires, the bridge fetches the list of port/VLAN
216 * membership that has been modified by the AA.
218 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
219 static long long int aa_refresh_timer = LLONG_MIN;
221 /* Whenever system interfaces are added, removed or change state, the bridge
222 * will be reconfigured.
224 static struct if_notifier *ifnotifier;
225 static bool ifaces_changed = false;
227 static void add_del_bridges(const struct ovsrec_open_vswitch *);
228 static void bridge_run__(void);
229 static void bridge_create(const struct ovsrec_bridge *);
230 static void bridge_destroy(struct bridge *);
231 static struct bridge *bridge_lookup(const char *name);
232 static unixctl_cb_func bridge_unixctl_dump_flows;
233 static unixctl_cb_func bridge_unixctl_reconnect;
234 static size_t bridge_get_controllers(const struct bridge *br,
235 struct ovsrec_controller ***controllersp);
236 static void bridge_collect_wanted_ports(struct bridge *,
237 const unsigned long *splinter_vlans,
238 struct shash *wanted_ports);
239 static void bridge_delete_ofprotos(void);
240 static void bridge_delete_or_reconfigure_ports(struct bridge *);
241 static void bridge_del_ports(struct bridge *,
242 const struct shash *wanted_ports);
243 static void bridge_add_ports(struct bridge *,
244 const struct shash *wanted_ports);
246 static void bridge_configure_datapath_id(struct bridge *);
247 static void bridge_configure_netflow(struct bridge *);
248 static void bridge_configure_forward_bpdu(struct bridge *);
249 static void bridge_configure_mac_table(struct bridge *);
250 static void bridge_configure_mcast_snooping(struct bridge *);
251 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
252 static void bridge_configure_ipfix(struct bridge *);
253 static void bridge_configure_spanning_tree(struct bridge *);
254 static void bridge_configure_tables(struct bridge *);
255 static void bridge_configure_dp_desc(struct bridge *);
256 static void bridge_configure_aa(struct bridge *);
257 static void bridge_aa_refresh_queued(struct bridge *);
258 static bool bridge_aa_need_refresh(struct bridge *);
259 static void bridge_configure_remotes(struct bridge *,
260 const struct sockaddr_in *managers,
262 static void bridge_pick_local_hw_addr(struct bridge *,
263 uint8_t ea[ETH_ADDR_LEN],
264 struct iface **hw_addr_iface);
265 static uint64_t bridge_pick_datapath_id(struct bridge *,
266 const uint8_t bridge_ea[ETH_ADDR_LEN],
267 struct iface *hw_addr_iface);
268 static uint64_t dpid_from_hash(const void *, size_t nbytes);
269 static bool bridge_has_bond_fake_iface(const struct bridge *,
271 static bool port_is_bond_fake_iface(const struct port *);
273 static unixctl_cb_func qos_unixctl_show;
275 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
276 static void port_del_ifaces(struct port *);
277 static void port_destroy(struct port *);
278 static struct port *port_lookup(const struct bridge *, const char *name);
279 static void port_configure(struct port *);
280 static struct lacp_settings *port_configure_lacp(struct port *,
281 struct lacp_settings *);
282 static void port_configure_bond(struct port *, struct bond_settings *);
283 static bool port_is_synthetic(const struct port *);
285 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
286 static void run_system_stats(void);
288 static void bridge_configure_mirrors(struct bridge *);
289 static struct mirror *mirror_create(struct bridge *,
290 const struct ovsrec_mirror *);
291 static void mirror_destroy(struct mirror *);
292 static bool mirror_configure(struct mirror *);
293 static void mirror_refresh_stats(struct mirror *);
295 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
296 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
297 const struct ovsrec_port *);
298 static bool iface_is_internal(const struct ovsrec_interface *iface,
299 const struct ovsrec_bridge *br);
300 static const char *iface_get_type(const struct ovsrec_interface *,
301 const struct ovsrec_bridge *);
302 static void iface_destroy(struct iface *);
303 static void iface_destroy__(struct iface *);
304 static struct iface *iface_lookup(const struct bridge *, const char *name);
305 static struct iface *iface_find(const char *name);
306 static struct iface *iface_from_ofp_port(const struct bridge *,
307 ofp_port_t ofp_port);
308 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
309 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
310 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
311 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
312 static void iface_configure_cfm(struct iface *);
313 static void iface_refresh_cfm_stats(struct iface *);
314 static void iface_refresh_stats(struct iface *);
315 static void iface_refresh_netdev_status(struct iface *);
316 static void iface_refresh_ofproto_status(struct iface *);
317 static bool iface_is_synthetic(const struct iface *);
318 static ofp_port_t iface_get_requested_ofp_port(
319 const struct ovsrec_interface *);
320 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
323 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
325 * This is deprecated. It is only for compatibility with broken device drivers
326 * in old versions of Linux that do not properly support VLANs when VLAN
327 * devices are not used. When broken device drivers are no longer in
328 * widespread use, we will delete these interfaces. */
330 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
331 static bool vlan_splinters_enabled_anywhere;
333 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
334 static unsigned long int *collect_splinter_vlans(
335 const struct ovsrec_open_vswitch *);
336 static void configure_splinter_port(struct port *);
337 static void add_vlan_splinter_ports(struct bridge *,
338 const unsigned long int *splinter_vlans,
339 struct shash *ports);
341 static void discover_types(const struct ovsrec_open_vswitch *cfg);
344 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
346 struct shash iface_hints;
347 static bool initialized = false;
354 shash_init(&iface_hints);
357 for (i = 0; i < cfg->n_bridges; i++) {
358 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
361 for (j = 0; j < br_cfg->n_ports; j++) {
362 struct ovsrec_port *port_cfg = br_cfg->ports[j];
365 for (k = 0; k < port_cfg->n_interfaces; k++) {
366 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
367 struct iface_hint *iface_hint;
369 iface_hint = xmalloc(sizeof *iface_hint);
370 iface_hint->br_name = br_cfg->name;
371 iface_hint->br_type = br_cfg->datapath_type;
372 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
374 shash_add(&iface_hints, if_cfg->name, iface_hint);
380 ofproto_init(&iface_hints);
382 shash_destroy_free_data(&iface_hints);
387 if_change_cb(void *aux OVS_UNUSED)
389 ifaces_changed = true;
392 /* Public functions. */
394 /* Initializes the bridge module, configuring it to obtain its configuration
395 * from an OVSDB server accessed over 'remote', which should be a string in a
396 * form acceptable to ovsdb_idl_create(). */
398 bridge_init(const char *remote)
400 /* Create connection to database. */
401 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
402 idl_seqno = ovsdb_idl_get_seqno(idl);
403 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
404 ovsdb_idl_verify_write_only(idl);
406 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
407 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
408 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
409 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
410 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
411 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
412 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
413 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
414 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
416 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
417 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
418 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
419 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
420 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
421 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
422 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
426 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
427 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
428 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
429 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
430 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
431 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
446 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
447 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
448 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
449 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
450 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
451 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
452 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
454 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
455 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
456 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
457 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
459 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
461 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
463 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
464 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
466 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
467 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
468 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
469 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
471 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
472 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
473 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
474 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
475 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
477 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
479 /* Register unixctl commands. */
480 unixctl_command_register("qos/show", "interface", 1, 1,
481 qos_unixctl_show, NULL);
482 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
483 bridge_unixctl_dump_flows, NULL);
484 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
485 bridge_unixctl_reconnect, NULL);
494 ifnotifier = if_notifier_create(if_change_cb, NULL);
500 struct bridge *br, *next_br;
502 if_notifier_destroy(ifnotifier);
503 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
506 ovsdb_idl_destroy(idl);
509 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
510 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
511 * responsible for freeing '*managersp' (with free()).
513 * You may be asking yourself "why does ovs-vswitchd care?", because
514 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
515 * should not be and in fact is not directly involved in that. But
516 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
517 * it has to tell in-band control where the managers are to enable that.
518 * (Thus, only managers connected in-band and with non-loopback addresses
522 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
523 struct sockaddr_in **managersp, size_t *n_managersp)
525 struct sockaddr_in *managers = NULL;
526 size_t n_managers = 0;
530 /* Collect all of the potential targets from the "targets" columns of the
531 * rows pointed to by "manager_options", excluding any that are
534 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
535 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
537 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
538 sset_find_and_delete(&targets, m->target);
540 sset_add(&targets, m->target);
544 /* Now extract the targets' IP addresses. */
545 if (!sset_is_empty(&targets)) {
548 managers = xmalloc(sset_count(&targets) * sizeof *managers);
549 SSET_FOR_EACH (target, &targets) {
551 struct sockaddr_storage ss;
552 struct sockaddr_in in;
555 /* Ignore loopback. */
556 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
558 && sa.ss.ss_family == AF_INET
559 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
560 managers[n_managers++] = sa.in;
564 sset_destroy(&targets);
566 *managersp = managers;
567 *n_managersp = n_managers;
571 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
573 unsigned long int *splinter_vlans;
574 struct sockaddr_in *managers;
575 struct bridge *br, *next;
576 int sflow_bridge_number;
579 COVERAGE_INC(bridge_reconfigure);
581 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
582 OFPROTO_FLOW_LIMIT_DEFAULT));
583 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
584 OFPROTO_MAX_IDLE_DEFAULT));
585 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
587 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
590 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
591 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
593 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
594 * to 'ovs_cfg', with only very minimal configuration otherwise.
596 * This is mostly an update to bridge data structures. Nothing is pushed
597 * down to ofproto or lower layers. */
598 add_del_bridges(ovs_cfg);
599 splinter_vlans = collect_splinter_vlans(ovs_cfg);
600 HMAP_FOR_EACH (br, node, &all_bridges) {
601 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
602 bridge_del_ports(br, &br->wanted_ports);
604 free(splinter_vlans);
606 /* Start pushing configuration changes down to the ofproto layer:
608 * - Delete ofprotos that are no longer configured.
610 * - Delete ports that are no longer configured.
612 * - Reconfigure existing ports to their desired configurations, or
613 * delete them if not possible.
615 * We have to do all the deletions before we can do any additions, because
616 * the ports to be added might require resources that will be freed up by
617 * deletions (they might especially overlap in name). */
618 bridge_delete_ofprotos();
619 HMAP_FOR_EACH (br, node, &all_bridges) {
621 bridge_delete_or_reconfigure_ports(br);
625 /* Finish pushing configuration changes to the ofproto layer:
627 * - Create ofprotos that are missing.
629 * - Add ports that are missing. */
630 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
634 error = ofproto_create(br->name, br->type, &br->ofproto);
636 VLOG_ERR("failed to create bridge %s: %s", br->name,
637 ovs_strerror(error));
638 shash_destroy(&br->wanted_ports);
641 /* Trigger storing datapath version. */
642 seq_change(connectivity_seq_get());
646 HMAP_FOR_EACH (br, node, &all_bridges) {
647 bridge_add_ports(br, &br->wanted_ports);
648 shash_destroy(&br->wanted_ports);
651 reconfigure_system_stats(ovs_cfg);
653 /* Complete the configuration. */
654 sflow_bridge_number = 0;
655 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
656 HMAP_FOR_EACH (br, node, &all_bridges) {
659 /* We need the datapath ID early to allow LACP ports to use it as the
660 * default system ID. */
661 bridge_configure_datapath_id(br);
663 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
666 port_configure(port);
668 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
669 iface_set_ofport(iface->cfg, iface->ofp_port);
670 /* Clear eventual previous errors */
671 ovsrec_interface_set_error(iface->cfg, NULL);
672 iface_configure_cfm(iface);
673 iface_configure_qos(iface, port->cfg->qos);
674 iface_set_mac(br, port, iface);
675 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
677 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
681 bridge_configure_mirrors(br);
682 bridge_configure_forward_bpdu(br);
683 bridge_configure_mac_table(br);
684 bridge_configure_mcast_snooping(br);
685 bridge_configure_remotes(br, managers, n_managers);
686 bridge_configure_netflow(br);
687 bridge_configure_sflow(br, &sflow_bridge_number);
688 bridge_configure_ipfix(br);
689 bridge_configure_spanning_tree(br);
690 bridge_configure_tables(br);
691 bridge_configure_dp_desc(br);
692 bridge_configure_aa(br);
696 /* The ofproto-dpif provider does some final reconfiguration in its
697 * ->type_run() function. We have to call it before notifying the database
698 * client that reconfiguration is complete, otherwise there is a very
699 * narrow race window in which e.g. ofproto/trace will not recognize the
700 * new configuration (sometimes this causes unit test failures). */
704 /* Delete ofprotos which aren't configured or have the wrong type. Create
705 * ofprotos which don't exist but need to. */
707 bridge_delete_ofprotos(void)
714 /* Delete ofprotos with no bridge or with the wrong type. */
717 ofproto_enumerate_types(&types);
718 SSET_FOR_EACH (type, &types) {
721 ofproto_enumerate_names(type, &names);
722 SSET_FOR_EACH (name, &names) {
723 br = bridge_lookup(name);
724 if (!br || strcmp(type, br->type)) {
725 ofproto_delete(name, type);
729 sset_destroy(&names);
730 sset_destroy(&types);
734 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
736 if (*n >= *allocated) {
737 ports = x2nrealloc(ports, allocated, sizeof *ports);
739 ports[(*n)++] = port;
744 bridge_delete_or_reconfigure_ports(struct bridge *br)
746 struct ofproto_port ofproto_port;
747 struct ofproto_port_dump dump;
749 struct sset ofproto_ports;
750 struct port *port, *port_next;
752 /* List of "ofp_port"s to delete. We make a list instead of deleting them
753 * right away because ofproto implementations aren't necessarily able to
754 * iterate through a changing list of ports in an entirely robust way. */
761 sset_init(&ofproto_ports);
763 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
764 * that are not configured in the database. (This commonly happens when
765 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
767 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
768 * that have the wrong OpenFlow port number (and arrange to add them back
769 * with the correct OpenFlow port number). */
770 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
771 ofp_port_t requested_ofp_port;
774 sset_add(&ofproto_ports, ofproto_port.name);
776 iface = iface_lookup(br, ofproto_port.name);
778 /* No such iface is configured, so we should delete this
781 * As a corner case exception, keep the port if it's a bond fake
783 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
784 && !strcmp(ofproto_port.type, "internal")) {
790 if (strcmp(ofproto_port.type, iface->type)
791 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
792 /* The interface is the wrong type or can't be configured.
797 /* If the requested OpenFlow port for 'iface' changed, and it's not
798 * already the correct port, then we might want to temporarily delete
799 * this interface, so we can add it back again with the new OpenFlow
801 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
802 if (iface->ofp_port != OFPP_LOCAL &&
803 requested_ofp_port != OFPP_NONE &&
804 requested_ofp_port != iface->ofp_port) {
805 ofp_port_t victim_request;
806 struct iface *victim;
808 /* Check for an existing OpenFlow port currently occupying
809 * 'iface''s requested port number. If there isn't one, then
810 * delete this port. Otherwise we need to consider further. */
811 victim = iface_from_ofp_port(br, requested_ofp_port);
816 /* 'victim' is a port currently using 'iface''s requested port
817 * number. Unless 'victim' specifically requested that port
818 * number, too, then we can delete both 'iface' and 'victim'
819 * temporarily. (We'll add both of them back again later with new
820 * OpenFlow port numbers.)
822 * If 'victim' did request port number 'requested_ofp_port', just
823 * like 'iface', then that's a configuration inconsistency that we
824 * can't resolve. We might as well let it keep its current port
826 victim_request = iface_get_requested_ofp_port(victim->cfg);
827 if (victim_request != requested_ofp_port) {
828 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
829 iface_destroy(victim);
838 iface_destroy(iface);
839 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
841 for (i = 0; i < n; i++) {
842 ofproto_port_del(br->ofproto, del[i]);
846 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
847 * that we didn't see when we iterated through the datapath, i.e. ports
848 * that disappeared underneath use. This is an unusual situation, but it
849 * can happen in some cases:
851 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
852 * idea but could happen).
854 * - The port represented a device that disappeared, e.g. a tuntap
855 * device destroyed via "tunctl -d", a physical Ethernet device
856 * whose module was just unloaded via "rmmod", or a virtual NIC for a
857 * VM whose VM was just terminated. */
858 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
859 struct iface *iface, *iface_next;
861 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
862 if (!sset_contains(&ofproto_ports, iface->name)) {
863 iface_destroy__(iface);
867 if (list_is_empty(&port->ifaces)) {
871 sset_destroy(&ofproto_ports);
875 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
876 bool with_requested_port)
878 struct shash_node *port_node;
880 SHASH_FOR_EACH (port_node, wanted_ports) {
881 const struct ovsrec_port *port_cfg = port_node->data;
884 for (i = 0; i < port_cfg->n_interfaces; i++) {
885 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
886 ofp_port_t requested_ofp_port;
888 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
889 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
890 struct iface *iface = iface_lookup(br, iface_cfg->name);
893 iface_create(br, iface_cfg, port_cfg);
901 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
903 /* First add interfaces that request a particular port number. */
904 bridge_add_ports__(br, wanted_ports, true);
906 /* Then add interfaces that want automatic port number assignment.
907 * We add these afterward to avoid accidentally taking a specifically
908 * requested port number. */
909 bridge_add_ports__(br, wanted_ports, false);
913 port_configure(struct port *port)
915 const struct ovsrec_port *cfg = port->cfg;
916 struct bond_settings bond_settings;
917 struct lacp_settings lacp_settings;
918 struct ofproto_bundle_settings s;
921 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
922 configure_splinter_port(port);
931 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
932 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
933 s.slaves[s.n_slaves++] = iface->ofp_port;
938 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
942 /* Get VLAN trunks. */
945 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
949 if (cfg->vlan_mode) {
950 if (!strcmp(cfg->vlan_mode, "access")) {
951 s.vlan_mode = PORT_VLAN_ACCESS;
952 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
953 s.vlan_mode = PORT_VLAN_TRUNK;
954 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
955 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
956 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
957 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
959 /* This "can't happen" because ovsdb-server should prevent it. */
960 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
961 "back to trunk mode", port->name, cfg->vlan_mode);
962 s.vlan_mode = PORT_VLAN_TRUNK;
966 s.vlan_mode = PORT_VLAN_ACCESS;
968 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
972 s.vlan_mode = PORT_VLAN_TRUNK;
975 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
978 /* Get LACP settings. */
979 s.lacp = port_configure_lacp(port, &lacp_settings);
983 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
984 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
985 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
988 s.lacp_slaves = NULL;
991 /* Get bond settings. */
992 if (s.n_slaves > 1) {
993 s.bond = &bond_settings;
994 port_configure_bond(port, &bond_settings);
997 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
998 netdev_set_miimon_interval(iface->netdev, 0);
1003 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1008 free(s.lacp_slaves);
1011 /* Pick local port hardware address and datapath ID for 'br'. */
1013 bridge_configure_datapath_id(struct bridge *br)
1015 uint8_t ea[ETH_ADDR_LEN];
1017 struct iface *local_iface;
1018 struct iface *hw_addr_iface;
1021 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1022 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1024 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1026 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1027 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1028 "Ethernet address: %s",
1029 br->name, ovs_strerror(error));
1032 memcpy(br->ea, ea, ETH_ADDR_LEN);
1034 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1035 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1036 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1037 ofproto_set_datapath_id(br->ofproto, dpid);
1040 dpid_string = xasprintf("%016"PRIx64, dpid);
1041 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1045 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1048 bridge_get_allowed_versions(struct bridge *br)
1050 if (!br->cfg->n_protocols)
1053 return ofputil_versions_from_strings(br->cfg->protocols,
1054 br->cfg->n_protocols);
1057 /* Set NetFlow configuration on 'br'. */
1059 bridge_configure_netflow(struct bridge *br)
1061 struct ovsrec_netflow *cfg = br->cfg->netflow;
1062 struct netflow_options opts;
1065 ofproto_set_netflow(br->ofproto, NULL);
1069 memset(&opts, 0, sizeof opts);
1071 /* Get default NetFlow configuration from datapath.
1072 * Apply overrides from 'cfg'. */
1073 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1074 if (cfg->engine_type) {
1075 opts.engine_type = *cfg->engine_type;
1077 if (cfg->engine_id) {
1078 opts.engine_id = *cfg->engine_id;
1081 /* Configure active timeout interval. */
1082 opts.active_timeout = cfg->active_timeout;
1083 if (!opts.active_timeout) {
1084 opts.active_timeout = -1;
1085 } else if (opts.active_timeout < 0) {
1086 VLOG_WARN("bridge %s: active timeout interval set to negative "
1087 "value, using default instead (%d seconds)", br->name,
1088 NF_ACTIVE_TIMEOUT_DEFAULT);
1089 opts.active_timeout = -1;
1092 /* Add engine ID to interface number to disambiguate bridgs? */
1093 opts.add_id_to_iface = cfg->add_id_to_interface;
1094 if (opts.add_id_to_iface) {
1095 if (opts.engine_id > 0x7f) {
1096 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1097 "another vswitch, choose an engine id less than 128",
1100 if (hmap_count(&br->ports) > 508) {
1101 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1102 "another port when more than 508 ports are used",
1108 sset_init(&opts.collectors);
1109 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1112 if (ofproto_set_netflow(br->ofproto, &opts)) {
1113 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1115 sset_destroy(&opts.collectors);
1118 /* Set sFlow configuration on 'br'. */
1120 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1122 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1123 struct ovsrec_controller **controllers;
1124 struct ofproto_sflow_options oso;
1125 size_t n_controllers;
1129 ofproto_set_sflow(br->ofproto, NULL);
1133 memset(&oso, 0, sizeof oso);
1135 sset_init(&oso.targets);
1136 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1138 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1139 if (cfg->sampling) {
1140 oso.sampling_rate = *cfg->sampling;
1143 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1145 oso.polling_interval = *cfg->polling;
1148 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1150 oso.header_len = *cfg->header;
1153 oso.sub_id = (*sflow_bridge_number)++;
1154 oso.agent_device = cfg->agent;
1156 oso.control_ip = NULL;
1157 n_controllers = bridge_get_controllers(br, &controllers);
1158 for (i = 0; i < n_controllers; i++) {
1159 if (controllers[i]->local_ip) {
1160 oso.control_ip = controllers[i]->local_ip;
1164 ofproto_set_sflow(br->ofproto, &oso);
1166 sset_destroy(&oso.targets);
1169 /* Returns whether a IPFIX row is valid. */
1171 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1173 return ipfix && ipfix->n_targets > 0;
1176 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1178 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1179 const struct bridge *br)
1181 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1184 /* Set IPFIX configuration on 'br'. */
1186 bridge_configure_ipfix(struct bridge *br)
1188 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1189 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1190 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1191 struct ofproto_ipfix_bridge_exporter_options be_opts;
1192 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1193 size_t n_fe_opts = 0;
1195 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1196 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1201 if (!valid_be_cfg && n_fe_opts == 0) {
1202 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1207 memset(&be_opts, 0, sizeof be_opts);
1209 sset_init(&be_opts.targets);
1210 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1212 if (be_cfg->sampling) {
1213 be_opts.sampling_rate = *be_cfg->sampling;
1215 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1217 if (be_cfg->obs_domain_id) {
1218 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1220 if (be_cfg->obs_point_id) {
1221 be_opts.obs_point_id = *be_cfg->obs_point_id;
1223 if (be_cfg->cache_active_timeout) {
1224 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1226 if (be_cfg->cache_max_flows) {
1227 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1230 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1231 "enable-tunnel-sampling", true);
1233 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1234 "enable-input-sampling", false);
1236 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1237 "enable-output-sampling", false);
1240 if (n_fe_opts > 0) {
1241 struct ofproto_ipfix_flow_exporter_options *opts;
1242 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1244 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1245 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1246 opts->collector_set_id = fe_cfg->id;
1247 sset_init(&opts->targets);
1248 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1249 fe_cfg->ipfix->n_targets);
1250 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1251 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1252 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1253 ? *fe_cfg->ipfix->cache_max_flows : 0;
1259 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1263 sset_destroy(&be_opts.targets);
1266 if (n_fe_opts > 0) {
1267 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1269 for (i = 0; i < n_fe_opts; i++) {
1270 sset_destroy(&opts->targets);
1278 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1279 struct ofproto_port_stp_settings *port_s,
1280 int *port_num_counter, unsigned long *port_num_bitmap)
1282 const char *config_str;
1283 struct iface *iface;
1285 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1286 port_s->enable = false;
1289 port_s->enable = true;
1292 /* STP over bonds is not supported. */
1293 if (!list_is_singleton(&port->ifaces)) {
1294 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1296 port_s->enable = false;
1300 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1302 /* Internal ports shouldn't participate in spanning tree, so
1304 if (!strcmp(iface->type, "internal")) {
1305 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1306 port_s->enable = false;
1310 /* STP on mirror output ports is not supported. */
1311 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1312 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1313 port_s->enable = false;
1317 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1319 unsigned long int port_num = strtoul(config_str, NULL, 0);
1320 int port_idx = port_num - 1;
1322 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1323 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1324 port_s->enable = false;
1328 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1329 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1330 port->name, port_num);
1331 port_s->enable = false;
1334 bitmap_set1(port_num_bitmap, port_idx);
1335 port_s->port_num = port_idx;
1337 if (*port_num_counter >= STP_MAX_PORTS) {
1338 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1339 port_s->enable = false;
1343 port_s->port_num = (*port_num_counter)++;
1346 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1348 port_s->path_cost = strtoul(config_str, NULL, 10);
1350 enum netdev_features current;
1353 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1354 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1355 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1358 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1360 port_s->priority = strtoul(config_str, NULL, 0);
1362 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1367 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1368 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1370 const char *config_str;
1371 struct iface *iface;
1373 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1374 port_s->enable = false;
1377 port_s->enable = true;
1380 /* RSTP over bonds is not supported. */
1381 if (!list_is_singleton(&port->ifaces)) {
1382 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1384 port_s->enable = false;
1388 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1390 /* Internal ports shouldn't participate in spanning tree, so
1392 if (!strcmp(iface->type, "internal")) {
1393 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1394 port_s->enable = false;
1398 /* RSTP on mirror output ports is not supported. */
1399 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1400 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1401 port_s->enable = false;
1405 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1407 unsigned long int port_num = strtoul(config_str, NULL, 0);
1408 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1409 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1410 port_s->enable = false;
1413 port_s->port_num = port_num;
1415 if (*port_num_counter >= RSTP_MAX_PORTS) {
1416 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1417 port_s->enable = false;
1420 /* If rstp-port-num is not specified, use 0.
1421 * rstp_port_set_port_number() will look for the first free one. */
1422 port_s->port_num = 0;
1425 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1427 port_s->path_cost = strtoul(config_str, NULL, 10);
1429 enum netdev_features current;
1432 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1433 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1434 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1437 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1439 port_s->priority = strtoul(config_str, NULL, 0);
1441 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1444 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1446 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1448 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1451 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1452 "rstp-admin-port-state", true);
1454 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1455 "rstp-port-admin-edge", false);
1456 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1457 "rstp-port-auto-edge", true);
1458 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1459 "rstp-port-mcheck", false);
1462 /* Set spanning tree configuration on 'br'. */
1464 bridge_configure_stp(struct bridge *br, bool enable_stp)
1467 ofproto_set_stp(br->ofproto, NULL);
1469 struct ofproto_stp_settings br_s;
1470 const char *config_str;
1472 int port_num_counter;
1473 unsigned long *port_num_bitmap;
1475 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1477 uint8_t ea[ETH_ADDR_LEN];
1479 if (eth_addr_from_string(config_str, ea)) {
1480 br_s.system_id = eth_addr_to_uint64(ea);
1482 br_s.system_id = eth_addr_to_uint64(br->ea);
1483 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1484 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1487 br_s.system_id = eth_addr_to_uint64(br->ea);
1490 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1492 br_s.priority = strtoul(config_str, NULL, 0);
1494 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1497 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1499 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1501 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1504 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1506 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1508 br_s.max_age = STP_DEFAULT_MAX_AGE;
1511 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1513 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1515 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1518 /* Configure STP on the bridge. */
1519 if (ofproto_set_stp(br->ofproto, &br_s)) {
1520 VLOG_ERR("bridge %s: could not enable STP", br->name);
1524 /* Users must either set the port number with the "stp-port-num"
1525 * configuration on all ports or none. If manual configuration
1526 * is not done, then we allocate them sequentially. */
1527 port_num_counter = 0;
1528 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1529 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1530 struct ofproto_port_stp_settings port_s;
1531 struct iface *iface;
1533 port_configure_stp(br->ofproto, port, &port_s,
1534 &port_num_counter, port_num_bitmap);
1536 /* As bonds are not supported, just apply configuration to
1537 * all interfaces. */
1538 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1539 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1541 VLOG_ERR("port %s: could not enable STP", port->name);
1547 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1548 && port_num_counter) {
1549 VLOG_ERR("bridge %s: must manually configure all STP port "
1550 "IDs or none, disabling", br->name);
1551 ofproto_set_stp(br->ofproto, NULL);
1553 bitmap_free(port_num_bitmap);
1558 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1561 ofproto_set_rstp(br->ofproto, NULL);
1563 struct ofproto_rstp_settings br_s;
1564 const char *config_str;
1566 int port_num_counter;
1568 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1570 uint8_t ea[ETH_ADDR_LEN];
1572 if (eth_addr_from_string(config_str, ea)) {
1573 br_s.address = eth_addr_to_uint64(ea);
1576 br_s.address = eth_addr_to_uint64(br->ea);
1577 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1578 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1582 br_s.address = eth_addr_to_uint64(br->ea);
1585 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1587 br_s.priority = strtoul(config_str, NULL, 0);
1589 br_s.priority = RSTP_DEFAULT_PRIORITY;
1592 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1594 br_s.ageing_time = strtoul(config_str, NULL, 0);
1596 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1599 config_str = smap_get(&br->cfg->other_config,
1600 "rstp-force-protocol-version");
1602 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1604 br_s.force_protocol_version = FPV_DEFAULT;
1607 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1609 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1611 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1614 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1616 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1618 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1621 config_str = smap_get(&br->cfg->other_config,
1622 "rstp-transmit-hold-count");
1624 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1626 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1629 /* Configure RSTP on the bridge. */
1630 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1631 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1635 port_num_counter = 0;
1636 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1637 struct ofproto_port_rstp_settings port_s;
1638 struct iface *iface;
1640 port_configure_rstp(br->ofproto, port, &port_s,
1643 /* As bonds are not supported, just apply configuration to
1644 * all interfaces. */
1645 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1646 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1648 VLOG_ERR("port %s: could not enable RSTP", port->name);
1657 bridge_configure_spanning_tree(struct bridge *br)
1659 bool enable_rstp = br->cfg->rstp_enable;
1660 bool enable_stp = br->cfg->stp_enable;
1662 if (enable_rstp && enable_stp) {
1663 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1664 "configured; enabling RSTP", br->name);
1668 bridge_configure_stp(br, enable_stp);
1669 bridge_configure_rstp(br, enable_rstp);
1673 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1675 const struct port *port = port_lookup(br, name);
1676 return port && port_is_bond_fake_iface(port);
1680 port_is_bond_fake_iface(const struct port *port)
1682 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1686 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1688 struct bridge *br, *next;
1689 struct shash new_br;
1692 /* Collect new bridges' names and types. */
1693 shash_init(&new_br);
1694 for (i = 0; i < cfg->n_bridges; i++) {
1695 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1696 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1698 if (strchr(br_cfg->name, '/')) {
1699 /* Prevent remote ovsdb-server users from accessing arbitrary
1700 * directories, e.g. consider a bridge named "../../../etc/". */
1701 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1703 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1704 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1708 /* Get rid of deleted bridges or those whose types have changed.
1709 * Update 'cfg' of bridges that still exist. */
1710 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1711 br->cfg = shash_find_data(&new_br, br->name);
1712 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1713 br->cfg->datapath_type))) {
1718 /* Add new bridges. */
1719 for (i = 0; i < cfg->n_bridges; i++) {
1720 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1721 struct bridge *br = bridge_lookup(br_cfg->name);
1723 bridge_create(br_cfg);
1727 shash_destroy(&new_br);
1730 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1731 * Returns 0 if successful, otherwise a positive errno value. */
1733 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1734 struct netdev *netdev, char **errp)
1736 return netdev_set_config(netdev, &iface_cfg->options, errp);
1739 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1740 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1742 * If successful, returns 0 and stores the network device in '*netdevp'. On
1743 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1745 iface_do_create(const struct bridge *br,
1746 const struct ovsrec_interface *iface_cfg,
1747 const struct ovsrec_port *port_cfg,
1748 ofp_port_t *ofp_portp, struct netdev **netdevp,
1751 struct netdev *netdev = NULL;
1754 if (netdev_is_reserved_name(iface_cfg->name)) {
1755 VLOG_WARN("could not create interface %s, name is reserved",
1761 error = netdev_open(iface_cfg->name,
1762 iface_get_type(iface_cfg, br->cfg), &netdev);
1764 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1765 iface_cfg->name, ovs_strerror(error));
1769 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1774 *ofp_portp = iface_pick_ofport(iface_cfg);
1775 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1780 VLOG_INFO("bridge %s: added interface %s on port %d",
1781 br->name, iface_cfg->name, *ofp_portp);
1783 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1784 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1792 netdev_close(netdev);
1796 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1797 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1798 * automatically allocated for the iface. Takes ownership of and
1799 * deallocates 'if_cfg'.
1801 * Return true if an iface is successfully created, false otherwise. */
1803 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1804 const struct ovsrec_port *port_cfg)
1806 struct netdev *netdev;
1807 struct iface *iface;
1808 ofp_port_t ofp_port;
1813 /* Do the bits that can fail up front. */
1814 ovs_assert(!iface_lookup(br, iface_cfg->name));
1815 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1817 iface_clear_db_record(iface_cfg, errp);
1822 /* Get or create the port structure. */
1823 port = port_lookup(br, port_cfg->name);
1825 port = port_create(br, port_cfg);
1828 /* Create the iface structure. */
1829 iface = xzalloc(sizeof *iface);
1830 list_push_back(&port->ifaces, &iface->port_elem);
1831 hmap_insert(&br->iface_by_name, &iface->name_node,
1832 hash_string(iface_cfg->name, 0));
1834 iface->name = xstrdup(iface_cfg->name);
1835 iface->ofp_port = ofp_port;
1836 iface->netdev = netdev;
1837 iface->type = iface_get_type(iface_cfg, br->cfg);
1838 iface->cfg = iface_cfg;
1839 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1840 hash_ofp_port(ofp_port));
1842 /* Populate initial status in database. */
1843 iface_refresh_stats(iface);
1844 iface_refresh_netdev_status(iface);
1846 /* Add bond fake iface if necessary. */
1847 if (port_is_bond_fake_iface(port)) {
1848 struct ofproto_port ofproto_port;
1850 if (ofproto_port_query_by_name(br->ofproto, port->name,
1852 struct netdev *netdev;
1855 error = netdev_open(port->name, "internal", &netdev);
1857 ofp_port_t fake_ofp_port = OFPP_NONE;
1858 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1859 netdev_close(netdev);
1861 VLOG_WARN("could not open network device %s (%s)",
1862 port->name, ovs_strerror(error));
1865 /* Already exists, nothing to do. */
1866 ofproto_port_destroy(&ofproto_port);
1873 /* Set forward BPDU option. */
1875 bridge_configure_forward_bpdu(struct bridge *br)
1877 ofproto_set_forward_bpdu(br->ofproto,
1878 smap_get_bool(&br->cfg->other_config,
1883 /* Set MAC learning table configuration for 'br'. */
1885 bridge_configure_mac_table(struct bridge *br)
1887 const char *idle_time_str;
1890 const char *mac_table_size_str;
1893 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1894 idle_time = (idle_time_str && atoi(idle_time_str)
1895 ? atoi(idle_time_str)
1896 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1898 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1899 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1900 ? atoi(mac_table_size_str)
1903 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1906 /* Set multicast snooping table configuration for 'br'. */
1908 bridge_configure_mcast_snooping(struct bridge *br)
1910 if (!br->cfg->mcast_snooping_enable) {
1911 ofproto_set_mcast_snooping(br->ofproto, NULL);
1914 struct ofproto_mcast_snooping_settings br_s;
1915 const char *idle_time_str;
1916 const char *max_entries_str;
1918 idle_time_str = smap_get(&br->cfg->other_config,
1919 "mcast-snooping-aging-time");
1920 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1921 ? atoi(idle_time_str)
1922 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1924 max_entries_str = smap_get(&br->cfg->other_config,
1925 "mcast-snooping-table-size");
1926 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1927 ? atoi(max_entries_str)
1928 : MCAST_DEFAULT_MAX_ENTRIES);
1930 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1931 "mcast-snooping-disable-flood-unregistered",
1934 /* Configure multicast snooping on the bridge */
1935 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1936 VLOG_ERR("bridge %s: could not enable multicast snooping",
1941 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1942 struct ofproto_mcast_snooping_port_settings port_s;
1943 port_s.flood = smap_get_bool(&port->cfg->other_config,
1944 "mcast-snooping-flood", false);
1945 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1946 "mcast-snooping-flood-reports", false);
1947 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1948 VLOG_ERR("port %s: could not configure mcast snooping",
1956 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1957 const struct port *fake_br, struct iface **hw_addr_iface)
1959 struct hmapx mirror_output_ports;
1961 bool found_addr = false;
1965 /* Mirror output ports don't participate in picking the local hardware
1966 * address. ofproto can't help us find out whether a given port is a
1967 * mirror output because we haven't configured mirrors yet, so we need to
1968 * accumulate them ourselves. */
1969 hmapx_init(&mirror_output_ports);
1970 for (i = 0; i < br->cfg->n_mirrors; i++) {
1971 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1972 if (m->output_port) {
1973 hmapx_add(&mirror_output_ports, m->output_port);
1977 /* Otherwise choose the minimum non-local MAC address among all of the
1979 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1980 uint8_t iface_ea[ETH_ADDR_LEN];
1981 struct iface *candidate;
1982 struct iface *iface;
1984 /* Mirror output ports don't participate. */
1985 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1989 /* Choose the MAC address to represent the port. */
1991 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1992 /* Find the interface with this Ethernet address (if any) so that
1993 * we can provide the correct devname to the caller. */
1994 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1995 uint8_t candidate_ea[ETH_ADDR_LEN];
1996 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1997 && eth_addr_equals(iface_ea, candidate_ea)) {
2002 /* Choose the interface whose MAC address will represent the port.
2003 * The Linux kernel bonding code always chooses the MAC address of
2004 * the first slave added to a bond, and the Fedora networking
2005 * scripts always add slaves to a bond in alphabetical order, so
2006 * for compatibility we choose the interface with the name that is
2007 * first in alphabetical order. */
2008 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2009 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2014 /* The local port doesn't count (since we're trying to choose its
2015 * MAC address anyway). */
2016 if (iface->ofp_port == OFPP_LOCAL) {
2020 /* For fake bridges we only choose from ports with the same tag */
2021 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2022 if (!port->cfg->tag) {
2025 if (*port->cfg->tag != *fake_br->cfg->tag) {
2031 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2037 /* Compare against our current choice. */
2038 if (!eth_addr_is_multicast(iface_ea) &&
2039 !eth_addr_is_local(iface_ea) &&
2040 !eth_addr_is_reserved(iface_ea) &&
2041 !eth_addr_is_zero(iface_ea) &&
2042 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2044 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2045 *hw_addr_iface = iface;
2051 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2052 *hw_addr_iface = NULL;
2055 hmapx_destroy(&mirror_output_ports);
2059 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2060 struct iface **hw_addr_iface)
2063 *hw_addr_iface = NULL;
2065 /* Did the user request a particular MAC? */
2066 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2067 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2068 if (eth_addr_is_multicast(ea)) {
2069 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2070 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2071 } else if (eth_addr_is_zero(ea)) {
2072 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2078 /* Find a local hw address */
2079 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2082 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2083 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2084 * an interface on 'br', then that interface must be passed in as
2085 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2086 * 'hw_addr_iface' must be passed in as a null pointer. */
2088 bridge_pick_datapath_id(struct bridge *br,
2089 const uint8_t bridge_ea[ETH_ADDR_LEN],
2090 struct iface *hw_addr_iface)
2093 * The procedure for choosing a bridge MAC address will, in the most
2094 * ordinary case, also choose a unique MAC that we can use as a datapath
2095 * ID. In some special cases, though, multiple bridges will end up with
2096 * the same MAC address. This is OK for the bridges, but it will confuse
2097 * the OpenFlow controller, because each datapath needs a unique datapath
2100 * Datapath IDs must be unique. It is also very desirable that they be
2101 * stable from one run to the next, so that policy set on a datapath
2104 const char *datapath_id;
2107 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2108 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2112 if (!hw_addr_iface) {
2114 * A purely internal bridge, that is, one that has no non-virtual
2115 * network devices on it at all, is difficult because it has no
2116 * natural unique identifier at all.
2118 * When the host is a XenServer, we handle this case by hashing the
2119 * host's UUID with the name of the bridge. Names of bridges are
2120 * persistent across XenServer reboots, although they can be reused if
2121 * an internal network is destroyed and then a new one is later
2122 * created, so this is fairly effective.
2124 * When the host is not a XenServer, we punt by using a random MAC
2125 * address on each run.
2127 const char *host_uuid = xenserver_get_host_uuid();
2129 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2130 dpid = dpid_from_hash(combined, strlen(combined));
2136 return eth_addr_to_uint64(bridge_ea);
2140 dpid_from_hash(const void *data, size_t n)
2142 uint8_t hash[SHA1_DIGEST_SIZE];
2144 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2145 sha1_bytes(data, n, hash);
2146 eth_addr_mark_random(hash);
2147 return eth_addr_to_uint64(hash);
2151 iface_refresh_netdev_status(struct iface *iface)
2155 enum netdev_features current;
2156 enum netdev_flags flags;
2157 const char *link_state;
2158 uint8_t mac[ETH_ADDR_LEN];
2159 int64_t bps, mtu_64, ifindex64, link_resets;
2162 if (iface_is_synthetic(iface)) {
2166 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2167 && !status_txn_try_again) {
2171 iface->change_seq = netdev_get_change_seq(iface->netdev);
2175 if (!netdev_get_status(iface->netdev, &smap)) {
2176 ovsrec_interface_set_status(iface->cfg, &smap);
2178 ovsrec_interface_set_status(iface->cfg, NULL);
2181 smap_destroy(&smap);
2183 error = netdev_get_flags(iface->netdev, &flags);
2185 const char *state = flags & NETDEV_UP ? "up" : "down";
2187 ovsrec_interface_set_admin_state(iface->cfg, state);
2189 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2192 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2193 ovsrec_interface_set_link_state(iface->cfg, link_state);
2195 link_resets = netdev_get_carrier_resets(iface->netdev);
2196 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2198 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2199 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2201 ovsrec_interface_set_duplex(iface->cfg,
2202 netdev_features_is_full_duplex(current)
2204 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2206 ovsrec_interface_set_duplex(iface->cfg, NULL);
2207 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2210 error = netdev_get_mtu(iface->netdev, &mtu);
2213 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2215 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2218 error = netdev_get_etheraddr(iface->netdev, mac);
2220 char mac_string[32];
2222 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2223 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2225 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2228 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2229 * if there is no valid ifindex number. */
2230 ifindex64 = netdev_get_ifindex(iface->netdev);
2231 if (ifindex64 < 0) {
2234 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2238 iface_refresh_ofproto_status(struct iface *iface)
2242 if (iface_is_synthetic(iface)) {
2246 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2250 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2252 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2255 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2257 || status_txn_try_again) {
2258 iface_refresh_cfm_stats(iface);
2261 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2263 || status_txn_try_again) {
2267 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2268 iface->ofp_port, &smap);
2269 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2270 smap_destroy(&smap);
2274 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2277 iface_refresh_cfm_stats(struct iface *iface)
2279 const struct ovsrec_interface *cfg = iface->cfg;
2280 struct cfm_status status;
2283 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2284 iface->ofp_port, &status);
2286 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2287 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2288 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2289 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2290 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2291 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2293 const char *reasons[CFM_FAULT_N_REASONS];
2294 int64_t cfm_health = status.health;
2295 int64_t cfm_flap_count = status.flap_count;
2296 bool faulted = status.faults != 0;
2299 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2302 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2303 int reason = 1 << i;
2304 if (status.faults & reason) {
2305 reasons[j++] = cfm_fault_reason_to_str(reason);
2308 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2310 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2312 if (status.remote_opstate >= 0) {
2313 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2314 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2316 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2319 ovsrec_interface_set_cfm_remote_mpids(cfg,
2320 (const int64_t *)status.rmps,
2322 if (cfm_health >= 0) {
2323 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2325 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2333 iface_refresh_stats(struct iface *iface)
2335 #define IFACE_STATS \
2336 IFACE_STAT(rx_packets, "rx_packets") \
2337 IFACE_STAT(tx_packets, "tx_packets") \
2338 IFACE_STAT(rx_bytes, "rx_bytes") \
2339 IFACE_STAT(tx_bytes, "tx_bytes") \
2340 IFACE_STAT(rx_dropped, "rx_dropped") \
2341 IFACE_STAT(tx_dropped, "tx_dropped") \
2342 IFACE_STAT(rx_errors, "rx_errors") \
2343 IFACE_STAT(tx_errors, "tx_errors") \
2344 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2345 IFACE_STAT(rx_over_errors, "rx_over_err") \
2346 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2347 IFACE_STAT(collisions, "collisions")
2349 #define IFACE_STAT(MEMBER, NAME) + 1
2350 enum { N_IFACE_STATS = IFACE_STATS };
2352 int64_t values[N_IFACE_STATS];
2353 const char *keys[N_IFACE_STATS];
2356 struct netdev_stats stats;
2358 if (iface_is_synthetic(iface)) {
2362 /* Intentionally ignore return value, since errors will set 'stats' to
2363 * all-1s, and we will deal with that correctly below. */
2364 netdev_get_stats(iface->netdev, &stats);
2366 /* Copy statistics into keys[] and values[]. */
2368 #define IFACE_STAT(MEMBER, NAME) \
2369 if (stats.MEMBER != UINT64_MAX) { \
2371 values[n] = stats.MEMBER; \
2376 ovs_assert(n <= N_IFACE_STATS);
2378 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2383 br_refresh_datapath_info(struct bridge *br)
2385 const char *version;
2387 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2388 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2391 ovsrec_bridge_set_datapath_version(br->cfg,
2392 version ? version : "<unknown>");
2396 br_refresh_stp_status(struct bridge *br)
2398 struct smap smap = SMAP_INITIALIZER(&smap);
2399 struct ofproto *ofproto = br->ofproto;
2400 struct ofproto_stp_status status;
2402 if (ofproto_get_stp_status(ofproto, &status)) {
2406 if (!status.enabled) {
2407 ovsrec_bridge_set_status(br->cfg, NULL);
2411 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2412 STP_ID_ARGS(status.bridge_id));
2413 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2414 STP_ID_ARGS(status.designated_root));
2415 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2417 ovsrec_bridge_set_status(br->cfg, &smap);
2418 smap_destroy(&smap);
2422 port_refresh_stp_status(struct port *port)
2424 struct ofproto *ofproto = port->bridge->ofproto;
2425 struct iface *iface;
2426 struct ofproto_port_stp_status status;
2429 if (port_is_synthetic(port)) {
2433 /* STP doesn't currently support bonds. */
2434 if (!list_is_singleton(&port->ifaces)) {
2435 ovsrec_port_set_status(port->cfg, NULL);
2439 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2440 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2444 if (!status.enabled) {
2445 ovsrec_port_set_status(port->cfg, NULL);
2449 /* Set Status column. */
2451 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2452 smap_add(&smap, "stp_state", stp_state_name(status.state));
2453 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2454 smap_add(&smap, "stp_role", stp_role_name(status.role));
2455 ovsrec_port_set_status(port->cfg, &smap);
2456 smap_destroy(&smap);
2460 port_refresh_stp_stats(struct port *port)
2462 struct ofproto *ofproto = port->bridge->ofproto;
2463 struct iface *iface;
2464 struct ofproto_port_stp_stats stats;
2465 const char *keys[3];
2466 int64_t int_values[3];
2468 if (port_is_synthetic(port)) {
2472 /* STP doesn't currently support bonds. */
2473 if (!list_is_singleton(&port->ifaces)) {
2477 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2478 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2482 if (!stats.enabled) {
2483 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2487 /* Set Statistics column. */
2488 keys[0] = "stp_tx_count";
2489 int_values[0] = stats.tx_count;
2490 keys[1] = "stp_rx_count";
2491 int_values[1] = stats.rx_count;
2492 keys[2] = "stp_error_count";
2493 int_values[2] = stats.error_count;
2495 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2496 ARRAY_SIZE(int_values));
2500 br_refresh_rstp_status(struct bridge *br)
2502 struct smap smap = SMAP_INITIALIZER(&smap);
2503 struct ofproto *ofproto = br->ofproto;
2504 struct ofproto_rstp_status status;
2506 if (ofproto_get_rstp_status(ofproto, &status)) {
2509 if (!status.enabled) {
2510 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2513 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2514 RSTP_ID_ARGS(status.bridge_id));
2515 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2516 status.root_path_cost);
2517 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2518 RSTP_ID_ARGS(status.root_id));
2519 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2520 RSTP_ID_ARGS(status.designated_id));
2521 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2522 status.designated_port_id);
2523 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2524 status.bridge_port_id);
2525 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2526 smap_destroy(&smap);
2530 port_refresh_rstp_status(struct port *port)
2532 struct ofproto *ofproto = port->bridge->ofproto;
2533 struct iface *iface;
2534 struct ofproto_port_rstp_status status;
2535 const char *keys[4];
2536 int64_t int_values[4];
2539 if (port_is_synthetic(port)) {
2543 /* RSTP doesn't currently support bonds. */
2544 if (!list_is_singleton(&port->ifaces)) {
2545 ovsrec_port_set_rstp_status(port->cfg, NULL);
2549 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2550 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2554 if (!status.enabled) {
2555 ovsrec_port_set_rstp_status(port->cfg, NULL);
2556 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2559 /* Set Status column. */
2562 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2564 smap_add_format(&smap, "rstp_port_role", "%s",
2565 rstp_port_role_name(status.role));
2566 smap_add_format(&smap, "rstp_port_state", "%s",
2567 rstp_state_name(status.state));
2568 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2569 RSTP_ID_ARGS(status.designated_bridge_id));
2570 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2571 status.designated_port_id);
2572 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2573 status.designated_path_cost);
2575 ovsrec_port_set_rstp_status(port->cfg, &smap);
2576 smap_destroy(&smap);
2578 /* Set Statistics column. */
2579 keys[0] = "rstp_tx_count";
2580 int_values[0] = status.tx_count;
2581 keys[1] = "rstp_rx_count";
2582 int_values[1] = status.rx_count;
2583 keys[2] = "rstp_uptime";
2584 int_values[2] = status.uptime;
2585 keys[3] = "rstp_error_count";
2586 int_values[3] = status.error_count;
2587 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2588 ARRAY_SIZE(int_values));
2592 port_refresh_bond_status(struct port *port, bool force_update)
2594 uint8_t mac[ETH_ADDR_LEN];
2596 /* Return if port is not a bond */
2597 if (list_is_singleton(&port->ifaces)) {
2601 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2605 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2606 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2612 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2614 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2618 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2620 bool enable = enable_system_stats(cfg);
2622 system_stats_enable(enable);
2624 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2629 run_system_stats(void)
2631 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2634 stats = system_stats_run();
2636 struct ovsdb_idl_txn *txn;
2637 struct ovsdb_datum datum;
2639 txn = ovsdb_idl_txn_create(idl);
2640 ovsdb_datum_from_smap(&datum, stats);
2641 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2643 ovsdb_idl_txn_commit(txn);
2644 ovsdb_idl_txn_destroy(txn);
2651 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2654 case OFPCR12_ROLE_EQUAL:
2656 case OFPCR12_ROLE_MASTER:
2658 case OFPCR12_ROLE_SLAVE:
2660 case OFPCR12_ROLE_NOCHANGE:
2662 return "*** INVALID ROLE ***";
2667 refresh_controller_status(void)
2671 const struct ovsrec_controller *cfg;
2675 /* Accumulate status for controllers on all bridges. */
2676 HMAP_FOR_EACH (br, node, &all_bridges) {
2677 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2680 /* Update each controller in the database with current status. */
2681 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2682 struct ofproto_controller_info *cinfo =
2683 shash_find_data(&info, cfg->target);
2686 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2687 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2689 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2691 ovsrec_controller_set_is_connected(cfg, false);
2692 ovsrec_controller_set_role(cfg, NULL);
2693 ovsrec_controller_set_status(cfg, NULL);
2697 ofproto_free_ofproto_controller_info(&info);
2700 /* Update interface and mirror statistics if necessary. */
2702 run_stats_update(void)
2704 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2711 /* Statistics update interval should always be greater than or equal to
2713 stats_interval = MAX(smap_get_int(&cfg->other_config,
2714 "stats-update-interval",
2716 if (stats_timer_interval != stats_interval) {
2717 stats_timer_interval = stats_interval;
2718 stats_timer = LLONG_MIN;
2721 if (time_msec() >= stats_timer) {
2722 enum ovsdb_idl_txn_status status;
2724 /* Rate limit the update. Do not start a new update if the
2725 * previous one is not done. */
2729 stats_txn = ovsdb_idl_txn_create(idl);
2730 HMAP_FOR_EACH (br, node, &all_bridges) {
2734 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2735 struct iface *iface;
2737 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2738 iface_refresh_stats(iface);
2740 port_refresh_stp_stats(port);
2742 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2743 mirror_refresh_stats(m);
2746 refresh_controller_status();
2749 status = ovsdb_idl_txn_commit(stats_txn);
2750 if (status != TXN_INCOMPLETE) {
2751 stats_timer = time_msec() + stats_timer_interval;
2752 ovsdb_idl_txn_destroy(stats_txn);
2759 stats_update_wait(void)
2761 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2762 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2764 ovsdb_idl_txn_wait(stats_txn);
2766 poll_timer_wait_until(stats_timer);
2770 /* Update bridge/port/interface status if necessary. */
2772 run_status_update(void)
2777 /* Rate limit the update. Do not start a new update if the
2778 * previous one is not done. */
2779 seq = seq_read(connectivity_seq_get());
2780 if (seq != connectivity_seqno || status_txn_try_again) {
2783 connectivity_seqno = seq;
2784 status_txn = ovsdb_idl_txn_create(idl);
2785 HMAP_FOR_EACH (br, node, &all_bridges) {
2788 br_refresh_stp_status(br);
2789 br_refresh_rstp_status(br);
2790 br_refresh_datapath_info(br);
2791 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2792 struct iface *iface;
2794 port_refresh_stp_status(port);
2795 port_refresh_rstp_status(port);
2796 port_refresh_bond_status(port, status_txn_try_again);
2797 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2798 iface_refresh_netdev_status(iface);
2799 iface_refresh_ofproto_status(iface);
2806 /* Commit the transaction and get the status. If the transaction finishes,
2807 * then destroy the transaction. Otherwise, keep it so that we can check
2808 * progress the next time that this function is called. */
2810 enum ovsdb_idl_txn_status status;
2812 status = ovsdb_idl_txn_commit(status_txn);
2813 if (status != TXN_INCOMPLETE) {
2814 ovsdb_idl_txn_destroy(status_txn);
2817 /* Sets the 'status_txn_try_again' if the transaction fails. */
2818 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2819 status_txn_try_again = false;
2821 status_txn_try_again = true;
2826 /* Refresh AA port status if necessary. */
2827 if (time_msec() >= aa_refresh_timer) {
2830 HMAP_FOR_EACH (br, node, &all_bridges) {
2831 if (bridge_aa_need_refresh(br)) {
2832 struct ovsdb_idl_txn *txn;
2834 txn = ovsdb_idl_txn_create(idl);
2835 bridge_aa_refresh_queued(br);
2836 ovsdb_idl_txn_commit(txn);
2837 ovsdb_idl_txn_destroy(txn);
2841 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2846 status_update_wait(void)
2848 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2849 * transaction to complete. If the status update to database needs to be
2850 * run again (transaction fails), registers a timeout in
2851 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2852 * sequence number. */
2854 ovsdb_idl_txn_wait(status_txn);
2855 } else if (status_txn_try_again) {
2856 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2858 seq_wait(connectivity_seq_get(), connectivity_seqno);
2869 /* Let each datapath type do the work that it needs to do. */
2871 ofproto_enumerate_types(&types);
2872 SSET_FOR_EACH (type, &types) {
2873 ofproto_type_run(type);
2875 sset_destroy(&types);
2877 /* Let each bridge do the work that it needs to do. */
2878 HMAP_FOR_EACH (br, node, &all_bridges) {
2879 ofproto_run(br->ofproto);
2886 static struct ovsrec_open_vswitch null_cfg;
2887 const struct ovsrec_open_vswitch *cfg;
2889 bool vlan_splinters_changed;
2891 ovsrec_open_vswitch_init(&null_cfg);
2897 if (ovsdb_idl_is_lock_contended(idl)) {
2898 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2899 struct bridge *br, *next_br;
2901 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2902 "disabling this process (pid %ld) until it goes away",
2903 (long int) getpid());
2905 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2908 /* Since we will not be running system_stats_run() in this process
2909 * with the current situation of multiple ovs-vswitchd daemons,
2910 * disable system stats collection. */
2911 system_stats_enable(false);
2913 } else if (!ovsdb_idl_has_lock(idl)
2914 || !ovsdb_idl_has_ever_connected(idl)) {
2915 /* Returns if not holding the lock or not done retrieving db
2919 cfg = ovsrec_open_vswitch_first(idl);
2921 /* Initialize the ofproto library. This only needs to run once, but
2922 * it must be done after the configuration is set. If the
2923 * initialization has already occurred, bridge_init_ofproto()
2924 * returns immediately. */
2925 bridge_init_ofproto(cfg);
2927 /* Once the value of flow-restore-wait is false, we no longer should
2928 * check its value from the database. */
2929 if (cfg && ofproto_get_flow_restore_wait()) {
2930 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2931 "flow-restore-wait", false));
2936 /* Re-configure SSL. We do this on every trip through the main loop,
2937 * instead of just when the database changes, because the contents of the
2938 * key and certificate files can change without the database changing.
2940 * We do this before bridge_reconfigure() because that function might
2941 * initiate SSL connections and thus requires SSL to be configured. */
2942 if (cfg && cfg->ssl) {
2943 const struct ovsrec_ssl *ssl = cfg->ssl;
2945 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2946 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2949 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2950 * usage has changed. */
2951 vlan_splinters_changed = false;
2952 if (vlan_splinters_enabled_anywhere) {
2955 HMAP_FOR_EACH (br, node, &all_bridges) {
2956 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2957 vlan_splinters_changed = true;
2963 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed
2964 || ifaces_changed) {
2965 struct ovsdb_idl_txn *txn;
2967 ifaces_changed = false;
2969 idl_seqno = ovsdb_idl_get_seqno(idl);
2970 txn = ovsdb_idl_txn_create(idl);
2971 bridge_reconfigure(cfg ? cfg : &null_cfg);
2974 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2975 discover_types(cfg);
2978 /* If we are completing our initial configuration for this run
2979 * of ovs-vswitchd, then keep the transaction around to monitor
2980 * it for completion. */
2981 if (initial_config_done) {
2982 /* Always sets the 'status_txn_try_again' to check again,
2983 * in case that this transaction fails. */
2984 status_txn_try_again = true;
2985 ovsdb_idl_txn_commit(txn);
2986 ovsdb_idl_txn_destroy(txn);
2988 initial_config_done = true;
2989 daemonize_txn = txn;
2993 if (daemonize_txn) {
2994 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2995 if (status != TXN_INCOMPLETE) {
2996 ovsdb_idl_txn_destroy(daemonize_txn);
2997 daemonize_txn = NULL;
2999 /* ovs-vswitchd has completed initialization, so allow the
3000 * process that forked us to exit successfully. */
3001 daemonize_complete();
3003 vlog_enable_async();
3005 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3010 run_status_update();
3020 ovsdb_idl_wait(idl);
3021 if (daemonize_txn) {
3022 ovsdb_idl_txn_wait(daemonize_txn);
3026 if (ifaces_changed) {
3027 poll_immediate_wake();
3031 ofproto_enumerate_types(&types);
3032 SSET_FOR_EACH (type, &types) {
3033 ofproto_type_wait(type);
3035 sset_destroy(&types);
3037 if (!hmap_is_empty(&all_bridges)) {
3040 HMAP_FOR_EACH (br, node, &all_bridges) {
3041 ofproto_wait(br->ofproto);
3043 stats_update_wait();
3044 status_update_wait();
3047 system_stats_wait();
3050 /* Adds some memory usage statistics for bridges into 'usage', for use with
3051 * memory_report(). */
3053 bridge_get_memory_usage(struct simap *usage)
3060 ofproto_enumerate_types(&types);
3061 SSET_FOR_EACH (type, &types) {
3062 ofproto_type_get_memory_usage(type, usage);
3064 sset_destroy(&types);
3066 HMAP_FOR_EACH (br, node, &all_bridges) {
3067 ofproto_get_memory_usage(br->ofproto, usage);
3071 /* QoS unixctl user interface functions. */
3073 struct qos_unixctl_show_cbdata {
3075 struct iface *iface;
3079 qos_unixctl_show_queue(unsigned int queue_id,
3080 const struct smap *details,
3081 struct iface *iface,
3084 struct netdev_queue_stats stats;
3085 struct smap_node *node;
3088 ds_put_cstr(ds, "\n");
3090 ds_put_format(ds, "Queue %u:\n", queue_id);
3092 ds_put_cstr(ds, "Default:\n");
3095 SMAP_FOR_EACH (node, details) {
3096 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3099 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3101 if (stats.tx_packets != UINT64_MAX) {
3102 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3105 if (stats.tx_bytes != UINT64_MAX) {
3106 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3109 if (stats.tx_errors != UINT64_MAX) {
3110 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3113 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3114 queue_id, ovs_strerror(error));
3119 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3120 const char *argv[], void *aux OVS_UNUSED)
3122 struct ds ds = DS_EMPTY_INITIALIZER;
3123 struct smap smap = SMAP_INITIALIZER(&smap);
3124 struct iface *iface;
3126 struct smap_node *node;
3128 iface = iface_find(argv[1]);
3130 unixctl_command_reply_error(conn, "no such interface");
3134 netdev_get_qos(iface->netdev, &type, &smap);
3136 if (*type != '\0') {
3137 struct netdev_queue_dump dump;
3138 struct smap details;
3139 unsigned int queue_id;
3141 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3143 SMAP_FOR_EACH (node, &smap) {
3144 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3147 smap_init(&details);
3148 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3149 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3151 smap_destroy(&details);
3153 unixctl_command_reply(conn, ds_cstr(&ds));
3155 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3156 unixctl_command_reply_error(conn, ds_cstr(&ds));
3159 smap_destroy(&smap);
3163 /* Bridge reconfiguration functions. */
3165 bridge_create(const struct ovsrec_bridge *br_cfg)
3169 ovs_assert(!bridge_lookup(br_cfg->name));
3170 br = xzalloc(sizeof *br);
3172 br->name = xstrdup(br_cfg->name);
3173 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3176 /* Derive the default Ethernet address from the bridge's UUID. This should
3177 * be unique and it will be stable between ovs-vswitchd runs. */
3178 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3179 eth_addr_mark_random(br->default_ea);
3181 hmap_init(&br->ports);
3182 hmap_init(&br->ifaces);
3183 hmap_init(&br->iface_by_name);
3184 hmap_init(&br->mirrors);
3186 hmap_init(&br->mappings);
3187 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3191 bridge_destroy(struct bridge *br)
3194 struct mirror *mirror, *next_mirror;
3195 struct port *port, *next_port;
3197 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3200 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3201 mirror_destroy(mirror);
3204 hmap_remove(&all_bridges, &br->node);
3205 ofproto_destroy(br->ofproto);
3206 hmap_destroy(&br->ifaces);
3207 hmap_destroy(&br->ports);
3208 hmap_destroy(&br->iface_by_name);
3209 hmap_destroy(&br->mirrors);
3210 hmap_destroy(&br->mappings);
3217 static struct bridge *
3218 bridge_lookup(const char *name)
3222 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3223 if (!strcmp(br->name, name)) {
3230 /* Handle requests for a listing of all flows known by the OpenFlow
3231 * stack, including those normally hidden. */
3233 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3234 const char *argv[], void *aux OVS_UNUSED)
3239 br = bridge_lookup(argv[1]);
3241 unixctl_command_reply_error(conn, "Unknown bridge");
3246 ofproto_get_all_flows(br->ofproto, &results);
3248 unixctl_command_reply(conn, ds_cstr(&results));
3249 ds_destroy(&results);
3252 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3253 * connections and reconnect. If BRIDGE is not specified, then all bridges
3254 * drop their controller connections and reconnect. */
3256 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3257 const char *argv[], void *aux OVS_UNUSED)
3261 br = bridge_lookup(argv[1]);
3263 unixctl_command_reply_error(conn, "Unknown bridge");
3266 ofproto_reconnect_controllers(br->ofproto);
3268 HMAP_FOR_EACH (br, node, &all_bridges) {
3269 ofproto_reconnect_controllers(br->ofproto);
3272 unixctl_command_reply(conn, NULL);
3276 bridge_get_controllers(const struct bridge *br,
3277 struct ovsrec_controller ***controllersp)
3279 struct ovsrec_controller **controllers;
3280 size_t n_controllers;
3282 controllers = br->cfg->controller;
3283 n_controllers = br->cfg->n_controller;
3285 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3291 *controllersp = controllers;
3293 return n_controllers;
3297 bridge_collect_wanted_ports(struct bridge *br,
3298 const unsigned long int *splinter_vlans,
3299 struct shash *wanted_ports)
3303 shash_init(wanted_ports);
3305 for (i = 0; i < br->cfg->n_ports; i++) {
3306 const char *name = br->cfg->ports[i]->name;
3307 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3308 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3312 if (bridge_get_controllers(br, NULL)
3313 && !shash_find(wanted_ports, br->name)) {
3314 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3315 br->name, br->name);
3317 ovsrec_interface_init(&br->synth_local_iface);
3318 ovsrec_port_init(&br->synth_local_port);
3320 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3321 br->synth_local_port.n_interfaces = 1;
3322 br->synth_local_port.name = br->name;
3324 br->synth_local_iface.name = br->name;
3325 br->synth_local_iface.type = "internal";
3327 br->synth_local_ifacep = &br->synth_local_iface;
3329 shash_add(wanted_ports, br->name, &br->synth_local_port);
3332 if (splinter_vlans) {
3333 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3337 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3338 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3339 * 'br' needs to complete its configuration. */
3341 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3343 struct shash_node *port_node;
3344 struct port *port, *next;
3346 /* Get rid of deleted ports.
3347 * Get rid of deleted interfaces on ports that still exist. */
3348 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3349 port->cfg = shash_find_data(wanted_ports, port->name);
3353 port_del_ifaces(port);
3357 /* Update iface->cfg and iface->type in interfaces that still exist. */
3358 SHASH_FOR_EACH (port_node, wanted_ports) {
3359 const struct ovsrec_port *port = port_node->data;
3362 for (i = 0; i < port->n_interfaces; i++) {
3363 const struct ovsrec_interface *cfg = port->interfaces[i];
3364 struct iface *iface = iface_lookup(br, cfg->name);
3365 const char *type = iface_get_type(cfg, br->cfg);
3370 } else if (!strcmp(type, "null")) {
3371 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3372 " may be removed in February 2013. Please email"
3373 " dev@openvswitch.org with concerns.",
3376 /* We will add new interfaces later. */
3382 /* Initializes 'oc' appropriately as a management service controller for
3385 * The caller must free oc->target when it is no longer needed. */
3387 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3388 struct ofproto_controller *oc)
3390 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3391 oc->max_backoff = 0;
3392 oc->probe_interval = 60;
3393 oc->band = OFPROTO_OUT_OF_BAND;
3395 oc->burst_limit = 0;
3396 oc->enable_async_msgs = true;
3400 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3402 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3403 struct ofproto_controller *oc)
3407 oc->target = c->target;
3408 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3409 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3410 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3411 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3412 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3413 oc->burst_limit = (c->controller_burst_limit
3414 ? *c->controller_burst_limit : 0);
3415 oc->enable_async_msgs = (!c->enable_async_messages
3416 || *c->enable_async_messages);
3417 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3418 if (dscp < 0 || dscp > 63) {
3419 dscp = DSCP_DEFAULT;
3424 /* Configures the IP stack for 'br''s local interface properly according to the
3425 * configuration in 'c'. */
3427 bridge_configure_local_iface_netdev(struct bridge *br,
3428 struct ovsrec_controller *c)
3430 struct netdev *netdev;
3431 struct in_addr mask, gateway;
3433 struct iface *local_iface;
3436 /* If there's no local interface or no IP address, give up. */
3437 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3438 if (!local_iface || !c->local_ip
3439 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3443 /* Bring up the local interface. */
3444 netdev = local_iface->netdev;
3445 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3447 /* Configure the IP address and netmask. */
3448 if (!c->local_netmask
3449 || !inet_pton(AF_INET, c->local_netmask, &mask)
3451 mask.s_addr = guess_netmask(ip.s_addr);
3453 if (!netdev_set_in4(netdev, ip, mask)) {
3454 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3455 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3458 /* Configure the default gateway. */
3459 if (c->local_gateway
3460 && inet_pton(AF_INET, c->local_gateway, &gateway)
3461 && gateway.s_addr) {
3462 if (!netdev_add_router(netdev, gateway)) {
3463 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3464 br->name, IP_ARGS(gateway.s_addr));
3469 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3470 * in either string are treated as equal to any number of slashes in the other,
3471 * e.g. "x///y" is equal to "x/y".
3473 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3474 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3475 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3476 * 'b' against a prefix of 'a'.
3479 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3481 const char *b_start = b;
3483 if (b - b_start >= b_stoplen) {
3485 } else if (*a != *b) {
3487 } else if (*a == '/') {
3488 a += strspn(a, "/");
3489 b += strspn(b, "/");
3490 } else if (*a == '\0') {
3500 bridge_configure_remotes(struct bridge *br,
3501 const struct sockaddr_in *managers, size_t n_managers)
3503 bool disable_in_band;
3505 struct ovsrec_controller **controllers;
3506 size_t n_controllers;
3508 enum ofproto_fail_mode fail_mode;
3510 struct ofproto_controller *ocs;
3514 /* Check if we should disable in-band control on this bridge. */
3515 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3518 /* Set OpenFlow queue ID for in-band control. */
3519 ofproto_set_in_band_queue(br->ofproto,
3520 smap_get_int(&br->cfg->other_config,
3521 "in-band-queue", -1));
3523 if (disable_in_band) {
3524 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3526 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3529 n_controllers = bridge_get_controllers(br, &controllers);
3531 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3534 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3535 for (i = 0; i < n_controllers; i++) {
3536 struct ovsrec_controller *c = controllers[i];
3538 if (!strncmp(c->target, "punix:", 6)
3539 || !strncmp(c->target, "unix:", 5)) {
3540 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3543 if (!strncmp(c->target, "unix:", 5)) {
3544 /* Connect to a listening socket */
3545 whitelist = xasprintf("unix:%s/", ovs_rundir());
3546 if (strchr(c->target, '/') &&
3547 !equal_pathnames(c->target, whitelist,
3548 strlen(whitelist))) {
3549 /* Absolute path specified, but not in ovs_rundir */
3550 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3551 "controller \"%s\" due to possibility for "
3552 "remote exploit. Instead, specify socket "
3553 "in whitelisted \"%s\" or connect to "
3554 "\"unix:%s/%s.mgmt\" (which is always "
3555 "available without special configuration).",
3556 br->name, c->target, whitelist,
3557 ovs_rundir(), br->name);
3562 whitelist = xasprintf("punix:%s/%s.controller",
3563 ovs_rundir(), br->name);
3564 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3565 /* Prevent remote ovsdb-server users from accessing
3566 * arbitrary Unix domain sockets and overwriting arbitrary
3568 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3569 "controller \"%s\" due to possibility of "
3570 "overwriting local files. Instead, specify "
3571 "whitelisted \"%s\" or connect to "
3572 "\"unix:%s/%s.mgmt\" (which is always "
3573 "available without special configuration).",
3574 br->name, c->target, whitelist,
3575 ovs_rundir(), br->name);
3584 bridge_configure_local_iface_netdev(br, c);
3585 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3586 if (disable_in_band) {
3587 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3592 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3593 bridge_get_allowed_versions(br));
3594 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3597 /* Set the fail-mode. */
3598 fail_mode = !br->cfg->fail_mode
3599 || !strcmp(br->cfg->fail_mode, "standalone")
3600 ? OFPROTO_FAIL_STANDALONE
3601 : OFPROTO_FAIL_SECURE;
3602 ofproto_set_fail_mode(br->ofproto, fail_mode);
3604 /* Configure OpenFlow controller connection snooping. */
3605 if (!ofproto_has_snoops(br->ofproto)) {
3609 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3610 ovs_rundir(), br->name));
3611 ofproto_set_snoops(br->ofproto, &snoops);
3612 sset_destroy(&snoops);
3617 bridge_configure_tables(struct bridge *br)
3619 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3623 n_tables = ofproto_get_n_tables(br->ofproto);
3625 for (i = 0; i < n_tables; i++) {
3626 struct ofproto_table_settings s;
3627 bool use_default_prefixes = true;
3630 s.max_flows = UINT_MAX;
3632 s.enable_eviction = false;
3634 s.n_prefix_fields = 0;
3635 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3637 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3638 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3641 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3642 s.max_flows = *cfg->flow_limit;
3645 s.enable_eviction = (cfg->overflow_policy
3646 && !strcmp(cfg->overflow_policy, "evict"));
3647 if (cfg->n_groups) {
3648 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3649 for (k = 0; k < cfg->n_groups; k++) {
3650 const char *string = cfg->groups[k];
3653 msg = mf_parse_subfield__(&s.groups[k], &string);
3655 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3656 "'groups' (%s)", br->name, i, msg);
3658 } else if (*string) {
3659 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3660 "element '%s' contains trailing garbage",
3661 br->name, i, cfg->groups[k]);
3668 /* Prefix lookup fields. */
3669 s.n_prefix_fields = 0;
3670 for (k = 0; k < cfg->n_prefixes; k++) {
3671 const char *name = cfg->prefixes[k];
3672 const struct mf_field *mf;
3674 if (strcmp(name, "none") == 0) {
3675 use_default_prefixes = false;
3676 s.n_prefix_fields = 0;
3679 mf = mf_from_name(name);
3681 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3685 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3686 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3687 "%s", br->name, name);
3690 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3691 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3692 "field not used: %s", br->name, name);
3695 use_default_prefixes = false;
3696 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3699 if (use_default_prefixes) {
3700 /* Use default values. */
3701 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3702 memcpy(s.prefix_fields, default_prefix_fields,
3703 sizeof default_prefix_fields);
3706 struct ds ds = DS_EMPTY_INITIALIZER;
3707 for (k = 0; k < s.n_prefix_fields; k++) {
3709 ds_put_char(&ds, ',');
3711 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3713 if (s.n_prefix_fields == 0) {
3714 ds_put_cstr(&ds, "none");
3716 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3717 br->name, i, ds_cstr(&ds));
3721 ofproto_configure_table(br->ofproto, i, &s);
3725 for (; j < br->cfg->n_flow_tables; j++) {
3726 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3727 "%"PRId64" not supported by this datapath", br->name,
3728 br->cfg->key_flow_tables[j]);
3733 bridge_configure_dp_desc(struct bridge *br)
3735 ofproto_set_dp_desc(br->ofproto,
3736 smap_get(&br->cfg->other_config, "dp-desc"));
3739 static struct aa_mapping *
3740 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3742 struct aa_mapping *m;
3744 HMAP_FOR_EACH_IN_BUCKET (m,
3746 hash_bytes(&isid, sizeof isid, 0),
3748 if (isid == m->isid) {
3755 static struct aa_mapping *
3756 bridge_aa_mapping_create(struct bridge *br,
3760 struct aa_mapping *m;
3762 m = xzalloc(sizeof *m);
3766 m->br_name = xstrdup(br->name);
3767 hmap_insert(&br->mappings,
3769 hash_bytes(&isid, sizeof isid, 0));
3775 bridge_aa_mapping_destroy(struct aa_mapping *m)
3778 struct bridge *br = m->bridge;
3781 ofproto_aa_mapping_unregister(br->ofproto, m);
3784 hmap_remove(&br->mappings, &m->hmap_node);
3793 bridge_aa_mapping_configure(struct aa_mapping *m)
3795 struct aa_mapping_settings s;
3801 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3807 bridge_configure_aa(struct bridge *br)
3809 const struct ovsdb_datum *mc;
3810 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3811 struct aa_settings aa_s;
3812 struct aa_mapping *m, *next;
3816 ofproto_set_aa(br->ofproto, NULL, NULL);
3820 memset(&aa_s, 0, sizeof aa_s);
3821 aa_s.system_description = auto_attach->system_description;
3822 aa_s.system_name = auto_attach->system_name;
3823 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3825 mc = ovsrec_autoattach_get_mappings(auto_attach,
3827 OVSDB_TYPE_INTEGER);
3828 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3829 union ovsdb_atom atom;
3831 atom.integer = m->isid;
3832 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3833 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3835 bridge_aa_mapping_destroy(m);
3839 /* Add new mappings and reconfigure existing ones. */
3840 for (i = 0; i < auto_attach->n_mappings; ++i) {
3841 struct aa_mapping *m =
3842 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3845 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3846 auto_attach->key_mappings[i],
3847 auto_attach->value_mappings[i]);
3848 m = bridge_aa_mapping_create(br,
3849 auto_attach->key_mappings[i],
3850 auto_attach->value_mappings[i]);
3852 if (!bridge_aa_mapping_configure(m)) {
3853 bridge_aa_mapping_destroy(m);
3860 bridge_aa_need_refresh(struct bridge *br)
3862 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3866 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3868 int64_t *trunks = NULL;
3870 bool found = false, reconfigure = false;
3872 for (i = 0; i < port->cfg->n_trunks; i++) {
3873 if (port->cfg->trunks[i] == m->vlan) {
3880 case BRIDGE_AA_VLAN_OPER_ADD:
3882 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3884 for (i = 0; i < port->cfg->n_trunks; i++) {
3885 trunks[i] = port->cfg->trunks[i];
3887 trunks[i++] = m->vlan;
3893 case BRIDGE_AA_VLAN_OPER_REMOVE:
3897 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3899 for (i = 0; i < port->cfg->n_trunks; i++) {
3900 if (port->cfg->trunks[i] != m->vlan) {
3901 trunks[j++] = port->cfg->trunks[i];
3910 case BRIDGE_AA_VLAN_OPER_UNDEF:
3912 VLOG_WARN("unrecognized operation %u", m->oper);
3917 /* VLAN switching under trunk mode cause the trunk port to switch all
3918 * VLANs, see ovs-vswitchd.conf.db
3921 static char *vlan_mode_access = "access";
3922 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3926 static char *vlan_mode_trunk = "trunk";
3927 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3930 ovsrec_port_set_trunks(port->cfg, trunks, i);
3932 /* Force reconfigure of the port. */
3933 port_configure(port);
3938 bridge_aa_refresh_queued(struct bridge *br)
3940 struct ovs_list *list = xmalloc(sizeof *list);
3941 struct bridge_aa_vlan *node, *next;
3944 ofproto_aa_vlan_get_queued(br->ofproto, list);
3946 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3949 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3952 port = port_lookup(br, node->port_name);
3954 bridge_aa_update_trunks(port, node);
3957 list_remove(&node->list_node);
3958 free(node->port_name);
3966 /* Port functions. */
3968 static struct port *
3969 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3973 port = xzalloc(sizeof *port);
3975 port->name = xstrdup(cfg->name);
3977 list_init(&port->ifaces);
3979 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3983 /* Deletes interfaces from 'port' that are no longer configured for it. */
3985 port_del_ifaces(struct port *port)
3987 struct iface *iface, *next;
3988 struct sset new_ifaces;
3991 /* Collect list of new interfaces. */
3992 sset_init(&new_ifaces);
3993 for (i = 0; i < port->cfg->n_interfaces; i++) {
3994 const char *name = port->cfg->interfaces[i]->name;
3995 const char *type = port->cfg->interfaces[i]->type;
3996 if (strcmp(type, "null")) {
3997 sset_add(&new_ifaces, name);
4001 /* Get rid of deleted interfaces. */
4002 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4003 if (!sset_contains(&new_ifaces, iface->name)) {
4004 iface_destroy(iface);
4008 sset_destroy(&new_ifaces);
4012 port_destroy(struct port *port)
4015 struct bridge *br = port->bridge;
4016 struct iface *iface, *next;
4019 ofproto_bundle_unregister(br->ofproto, port);
4022 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4023 iface_destroy__(iface);
4026 hmap_remove(&br->ports, &port->hmap_node);
4032 static struct port *
4033 port_lookup(const struct bridge *br, const char *name)
4037 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4039 if (!strcmp(port->name, name)) {
4047 enable_lacp(struct port *port, bool *activep)
4049 if (!port->cfg->lacp) {
4050 /* XXX when LACP implementation has been sufficiently tested, enable by
4051 * default and make active on bonded ports. */
4053 } else if (!strcmp(port->cfg->lacp, "off")) {
4055 } else if (!strcmp(port->cfg->lacp, "active")) {
4058 } else if (!strcmp(port->cfg->lacp, "passive")) {
4062 VLOG_WARN("port %s: unknown LACP mode %s",
4063 port->name, port->cfg->lacp);
4068 static struct lacp_settings *
4069 port_configure_lacp(struct port *port, struct lacp_settings *s)
4071 const char *lacp_time, *system_id;
4074 if (!enable_lacp(port, &s->active)) {
4078 s->name = port->name;
4080 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4082 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4083 ETH_ADDR_SCAN_ARGS(s->id))) {
4084 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4085 " address.", port->name, system_id);
4089 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4092 if (eth_addr_is_zero(s->id)) {
4093 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4097 /* Prefer bondable links if unspecified. */
4098 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4100 s->priority = (priority > 0 && priority <= UINT16_MAX
4102 : UINT16_MAX - !list_is_short(&port->ifaces));
4104 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4105 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4107 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4108 "lacp-fallback-ab", false);
4114 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4116 int priority, portid, key;
4118 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4119 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4121 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4123 if (portid <= 0 || portid > UINT16_MAX) {
4124 portid = ofp_to_u16(iface->ofp_port);
4127 if (priority <= 0 || priority > UINT16_MAX) {
4128 priority = UINT16_MAX;
4131 if (key < 0 || key > UINT16_MAX) {
4135 s->name = iface->name;
4137 s->priority = priority;
4142 port_configure_bond(struct port *port, struct bond_settings *s)
4144 const char *detect_s;
4145 struct iface *iface;
4147 int miimon_interval;
4149 s->name = port->name;
4151 if (port->cfg->bond_mode) {
4152 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4153 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4154 port->name, port->cfg->bond_mode,
4155 bond_mode_to_string(s->balance));
4158 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4160 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4161 * active-backup. At some point we should remove this warning. */
4162 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4163 " in previous versions, the default bond_mode was"
4164 " balance-slb", port->name,
4165 bond_mode_to_string(s->balance));
4167 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4168 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4169 "please use another bond type or disable flood_vlans",
4173 miimon_interval = smap_get_int(&port->cfg->other_config,
4174 "bond-miimon-interval", 0);
4175 if (miimon_interval <= 0) {
4176 miimon_interval = 200;
4179 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4180 if (!detect_s || !strcmp(detect_s, "carrier")) {
4181 miimon_interval = 0;
4182 } else if (strcmp(detect_s, "miimon")) {
4183 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4184 "defaulting to carrier", port->name, detect_s);
4185 miimon_interval = 0;
4188 s->up_delay = MAX(0, port->cfg->bond_updelay);
4189 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4190 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4191 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4192 "bond-rebalance-interval", 10000);
4193 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4194 s->rebalance_interval = 1000;
4197 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4198 "lacp-fallback-ab", false);
4200 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4201 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4204 mac_s = port->cfg->bond_active_slave;
4205 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4206 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4207 /* OVSDB did not store the last active interface */
4208 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4212 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4213 * instead of obtaining it from the database. */
4215 port_is_synthetic(const struct port *port)
4217 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4220 /* Interface functions. */
4223 iface_is_internal(const struct ovsrec_interface *iface,
4224 const struct ovsrec_bridge *br)
4226 /* The local port and "internal" ports are always "internal". */
4227 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4230 /* Returns the correct network device type for interface 'iface' in bridge
4233 iface_get_type(const struct ovsrec_interface *iface,
4234 const struct ovsrec_bridge *br)
4238 /* The local port always has type "internal". Other ports take
4239 * their type from the database and default to "system" if none is
4241 if (iface_is_internal(iface, br)) {
4244 type = iface->type[0] ? iface->type : "system";
4247 return ofproto_port_open_type(br->datapath_type, type);
4251 iface_destroy__(struct iface *iface)
4254 struct port *port = iface->port;
4255 struct bridge *br = port->bridge;
4257 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4258 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4261 if (iface->ofp_port != OFPP_NONE) {
4262 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4265 list_remove(&iface->port_elem);
4266 hmap_remove(&br->iface_by_name, &iface->name_node);
4268 /* The user is changing configuration here, so netdev_remove needs to be
4269 * used as opposed to netdev_close */
4270 netdev_remove(iface->netdev);
4278 iface_destroy(struct iface *iface)
4281 struct port *port = iface->port;
4283 iface_destroy__(iface);
4284 if (list_is_empty(&port->ifaces)) {
4290 static struct iface *
4291 iface_lookup(const struct bridge *br, const char *name)
4293 struct iface *iface;
4295 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4296 &br->iface_by_name) {
4297 if (!strcmp(iface->name, name)) {
4305 static struct iface *
4306 iface_find(const char *name)
4308 const struct bridge *br;
4310 HMAP_FOR_EACH (br, node, &all_bridges) {
4311 struct iface *iface = iface_lookup(br, name);
4320 static struct iface *
4321 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4323 struct iface *iface;
4325 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4327 if (iface->ofp_port == ofp_port) {
4334 /* Set Ethernet address of 'iface', if one is specified in the configuration
4337 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4339 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4340 struct iface *hw_addr_iface;
4342 if (strcmp(iface->type, "internal")) {
4346 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4348 } else if (port->cfg->fake_bridge) {
4349 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4350 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4355 if (iface->ofp_port == OFPP_LOCAL) {
4356 VLOG_ERR("interface %s: ignoring mac in Interface record "
4357 "(use Bridge record to set local port's mac)",
4359 } else if (eth_addr_is_multicast(mac)) {
4360 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4363 int error = netdev_set_etheraddr(iface->netdev, mac);
4365 VLOG_ERR("interface %s: setting MAC failed (%s)",
4366 iface->name, ovs_strerror(error));
4372 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4374 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4376 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4377 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4378 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4382 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4383 * sets the "ofport" field to -1.
4385 * This is appropriate when 'if_cfg''s interface cannot be created or is
4386 * otherwise invalid. */
4388 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4390 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4391 iface_set_ofport(if_cfg, OFPP_NONE);
4392 ovsrec_interface_set_error(if_cfg, errp);
4393 ovsrec_interface_set_status(if_cfg, NULL);
4394 ovsrec_interface_set_admin_state(if_cfg, NULL);
4395 ovsrec_interface_set_duplex(if_cfg, NULL);
4396 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4397 ovsrec_interface_set_link_state(if_cfg, NULL);
4398 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4399 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4400 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4401 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4402 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4403 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4404 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4405 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4410 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4412 union ovsdb_atom atom;
4414 atom.integer = target;
4415 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4419 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4421 struct ofpbuf queues_buf;
4423 ofpbuf_init(&queues_buf, 0);
4425 if (!qos || qos->type[0] == '\0') {
4426 netdev_set_qos(iface->netdev, NULL, NULL);
4428 const struct ovsdb_datum *queues;
4429 struct netdev_queue_dump dump;
4430 unsigned int queue_id;
4431 struct smap details;
4435 /* Configure top-level Qos for 'iface'. */
4436 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4438 /* Deconfigure queues that were deleted. */
4439 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4441 smap_init(&details);
4442 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4443 if (!queue_ids_include(queues, queue_id)) {
4444 netdev_delete_queue(iface->netdev, queue_id);
4447 smap_destroy(&details);
4449 /* Configure queues for 'iface'. */
4451 for (i = 0; i < qos->n_queues; i++) {
4452 const struct ovsrec_queue *queue = qos->value_queues[i];
4453 unsigned int queue_id = qos->key_queues[i];
4455 if (queue_id == 0) {
4459 if (queue->n_dscp == 1) {
4460 struct ofproto_port_queue *port_queue;
4462 port_queue = ofpbuf_put_uninit(&queues_buf,
4463 sizeof *port_queue);
4464 port_queue->queue = queue_id;
4465 port_queue->dscp = queue->dscp[0];
4468 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4471 struct smap details;
4473 smap_init(&details);
4474 netdev_set_queue(iface->netdev, 0, &details);
4475 smap_destroy(&details);
4479 if (iface->ofp_port != OFPP_NONE) {
4480 const struct ofproto_port_queue *port_queues = queues_buf.data;
4481 size_t n_queues = queues_buf.size / sizeof *port_queues;
4483 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4484 port_queues, n_queues);
4487 netdev_set_policing(iface->netdev,
4488 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4489 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4491 ofpbuf_uninit(&queues_buf);
4495 iface_configure_cfm(struct iface *iface)
4497 const struct ovsrec_interface *cfg = iface->cfg;
4498 const char *opstate_str;
4499 const char *cfm_ccm_vlan;
4500 struct cfm_settings s;
4501 struct smap netdev_args;
4503 if (!cfg->n_cfm_mpid) {
4504 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4508 s.check_tnl_key = false;
4509 smap_init(&netdev_args);
4510 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4511 const char *key = smap_get(&netdev_args, "key");
4512 const char *in_key = smap_get(&netdev_args, "in_key");
4514 s.check_tnl_key = (key && !strcmp(key, "flow"))
4515 || (in_key && !strcmp(in_key, "flow"));
4517 smap_destroy(&netdev_args);
4519 s.mpid = *cfg->cfm_mpid;
4520 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4521 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4522 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4524 if (s.interval <= 0) {
4528 if (!cfm_ccm_vlan) {
4530 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4531 s.ccm_vlan = CFM_RANDOM_VLAN;
4533 s.ccm_vlan = atoi(cfm_ccm_vlan);
4534 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4539 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4541 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4543 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4544 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4546 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4549 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4550 * instead of obtaining it from the database. */
4552 iface_is_synthetic(const struct iface *iface)
4554 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4558 iface_validate_ofport__(size_t n, int64_t *ofport)
4560 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4561 ? u16_to_ofp(*ofport)
4566 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4568 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4572 iface_pick_ofport(const struct ovsrec_interface *cfg)
4574 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4575 return (requested_ofport != OFPP_NONE
4577 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4580 /* Port mirroring. */
4582 static struct mirror *
4583 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4587 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4588 if (uuid_equals(uuid, &m->uuid)) {
4596 bridge_configure_mirrors(struct bridge *br)
4598 const struct ovsdb_datum *mc;
4599 unsigned long *flood_vlans;
4600 struct mirror *m, *next;
4603 /* Get rid of deleted mirrors. */
4604 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4605 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4606 union ovsdb_atom atom;
4608 atom.uuid = m->uuid;
4609 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4614 /* Add new mirrors and reconfigure existing ones. */
4615 for (i = 0; i < br->cfg->n_mirrors; i++) {
4616 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4617 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4619 m = mirror_create(br, cfg);
4622 if (!mirror_configure(m)) {
4627 /* Update flooded vlans (for RSPAN). */
4628 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4629 br->cfg->n_flood_vlans);
4630 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4631 bitmap_free(flood_vlans);
4634 static struct mirror *
4635 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4639 m = xzalloc(sizeof *m);
4640 m->uuid = cfg->header_.uuid;
4641 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4643 m->name = xstrdup(cfg->name);
4649 mirror_destroy(struct mirror *m)
4652 struct bridge *br = m->bridge;
4655 ofproto_mirror_unregister(br->ofproto, m);
4658 hmap_remove(&br->mirrors, &m->hmap_node);
4665 mirror_collect_ports(struct mirror *m,
4666 struct ovsrec_port **in_ports, int n_in_ports,
4667 void ***out_portsp, size_t *n_out_portsp)
4669 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4670 size_t n_out_ports = 0;
4673 for (i = 0; i < n_in_ports; i++) {
4674 const char *name = in_ports[i]->name;
4675 struct port *port = port_lookup(m->bridge, name);
4677 out_ports[n_out_ports++] = port;
4679 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4680 "port %s", m->bridge->name, m->name, name);
4683 *out_portsp = out_ports;
4684 *n_out_portsp = n_out_ports;
4688 mirror_configure(struct mirror *m)
4690 const struct ovsrec_mirror *cfg = m->cfg;
4691 struct ofproto_mirror_settings s;
4694 if (strcmp(cfg->name, m->name)) {
4696 m->name = xstrdup(cfg->name);
4700 /* Get output port or VLAN. */
4701 if (cfg->output_port) {
4702 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4703 if (!s.out_bundle) {
4704 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4705 m->bridge->name, m->name);
4708 s.out_vlan = UINT16_MAX;
4710 if (cfg->output_vlan) {
4711 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4712 "output vlan; ignoring output vlan",
4713 m->bridge->name, m->name);
4715 } else if (cfg->output_vlan) {
4716 /* The database should prevent invalid VLAN values. */
4717 s.out_bundle = NULL;
4718 s.out_vlan = *cfg->output_vlan;
4720 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4721 m->bridge->name, m->name);
4725 /* Get port selection. */
4726 if (cfg->select_all) {
4727 size_t n_ports = hmap_count(&m->bridge->ports);
4728 void **ports = xmalloc(n_ports * sizeof *ports);
4733 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4743 /* Get ports, dropping ports that don't exist.
4744 * The IDL ensures that there are no duplicates. */
4745 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4746 &s.srcs, &s.n_srcs);
4747 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4748 &s.dsts, &s.n_dsts);
4751 /* Get VLAN selection. */
4752 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4755 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4758 if (s.srcs != s.dsts) {
4767 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4769 * This is deprecated. It is only for compatibility with broken device drivers
4770 * in old versions of Linux that do not properly support VLANs when VLAN
4771 * devices are not used. When broken device drivers are no longer in
4772 * widespread use, we will delete these interfaces. */
4774 static struct ovsrec_port **recs;
4775 static size_t n_recs, allocated_recs;
4777 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4778 * splinters are reconfigured. */
4780 register_rec(struct ovsrec_port *rec)
4782 if (n_recs >= allocated_recs) {
4783 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4785 recs[n_recs++] = rec;
4788 /* Frees all of the ports registered with register_reg(). */
4790 free_registered_recs(void)
4794 for (i = 0; i < n_recs; i++) {
4795 struct ovsrec_port *port = recs[i];
4798 for (j = 0; j < port->n_interfaces; j++) {
4799 struct ovsrec_interface *iface = port->interfaces[j];
4804 smap_destroy(&port->other_config);
4805 free(port->interfaces);
4813 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4816 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4818 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4822 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4825 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4826 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4827 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4830 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4831 * use, returns NULL.
4833 * Updates 'vlan_splinters_enabled_anywhere'. */
4834 static unsigned long int *
4835 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4837 unsigned long int *splinter_vlans;
4838 struct sset splinter_ifaces;
4839 const char *real_dev_name;
4840 struct shash *real_devs;
4841 struct shash_node *node;
4845 /* Free space allocated for synthesized ports and interfaces, since we're
4846 * in the process of reconstructing all of them. */
4847 free_registered_recs();
4849 splinter_vlans = bitmap_allocate(4096);
4850 sset_init(&splinter_ifaces);
4851 vlan_splinters_enabled_anywhere = false;
4852 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4853 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4856 for (j = 0; j < br_cfg->n_ports; j++) {
4857 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4860 for (k = 0; k < port_cfg->n_interfaces; k++) {
4861 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4863 if (vlan_splinters_is_enabled(iface_cfg)) {
4864 vlan_splinters_enabled_anywhere = true;
4865 sset_add(&splinter_ifaces, iface_cfg->name);
4866 vlan_bitmap_from_array__(port_cfg->trunks,
4872 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4873 bitmap_set1(splinter_vlans, *port_cfg->tag);
4878 if (!vlan_splinters_enabled_anywhere) {
4879 free(splinter_vlans);
4880 sset_destroy(&splinter_ifaces);
4884 HMAP_FOR_EACH (br, node, &all_bridges) {
4886 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4890 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4891 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4892 * device to be created for it. */
4893 bitmap_set0(splinter_vlans, 0);
4894 bitmap_set0(splinter_vlans, 4095);
4896 /* Delete all VLAN devices that we don't need. */
4898 real_devs = vlandev_get_real_devs();
4899 SHASH_FOR_EACH (node, real_devs) {
4900 const struct vlan_real_dev *real_dev = node->data;
4901 const struct vlan_dev *vlan_dev;
4902 bool real_dev_has_splinters;
4904 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4906 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4907 if (!real_dev_has_splinters
4908 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4909 struct netdev *netdev;
4911 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4912 if (!netdev_get_in4(netdev, NULL, NULL) ||
4913 !netdev_get_in6(netdev, NULL)) {
4914 /* It has an IP address configured, so we don't own
4915 * it. Don't delete it. */
4917 vlandev_del(vlan_dev->name);
4919 netdev_close(netdev);
4926 /* Add all VLAN devices that we need. */
4927 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4930 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4931 if (!vlandev_get_name(real_dev_name, vid)) {
4932 vlandev_add(real_dev_name, vid);
4939 sset_destroy(&splinter_ifaces);
4941 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4942 free(splinter_vlans);
4945 return splinter_vlans;
4948 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4951 configure_splinter_port(struct port *port)
4953 struct ofproto *ofproto = port->bridge->ofproto;
4954 ofp_port_t realdev_ofp_port;
4955 const char *realdev_name;
4956 struct iface *vlandev, *realdev;
4958 ofproto_bundle_unregister(port->bridge->ofproto, port);
4960 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4963 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4964 realdev = iface_lookup(port->bridge, realdev_name);
4965 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4967 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4971 static struct ovsrec_port *
4972 synthesize_splinter_port(const char *real_dev_name,
4973 const char *vlan_dev_name, int vid)
4975 struct ovsrec_interface *iface;
4976 struct ovsrec_port *port;
4978 iface = xmalloc(sizeof *iface);
4979 ovsrec_interface_init(iface);
4980 iface->name = xstrdup(vlan_dev_name);
4981 iface->type = "system";
4983 port = xmalloc(sizeof *port);
4984 ovsrec_port_init(port);
4985 port->interfaces = xmemdup(&iface, sizeof iface);
4986 port->n_interfaces = 1;
4987 port->name = xstrdup(vlan_dev_name);
4988 port->vlan_mode = "splinter";
4989 port->tag = xmalloc(sizeof *port->tag);
4992 smap_add(&port->other_config, "realdev", real_dev_name);
4998 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4999 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
5000 * 1-bit in the 'splinter_vlans' bitmap. */
5002 add_vlan_splinter_ports(struct bridge *br,
5003 const unsigned long int *splinter_vlans,
5004 struct shash *ports)
5008 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
5009 * we're modifying 'ports'. */
5010 for (i = 0; i < br->cfg->n_ports; i++) {
5011 const char *name = br->cfg->ports[i]->name;
5012 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
5015 for (j = 0; j < port_cfg->n_interfaces; j++) {
5016 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
5018 if (vlan_splinters_is_enabled(iface_cfg)) {
5019 const char *real_dev_name;
5022 real_dev_name = iface_cfg->name;
5023 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
5024 const char *vlan_dev_name;
5026 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5028 && !shash_find(ports, vlan_dev_name)) {
5029 shash_add(ports, vlan_dev_name,
5030 synthesize_splinter_port(
5031 real_dev_name, vlan_dev_name, vid));
5040 mirror_refresh_stats(struct mirror *m)
5042 struct ofproto *ofproto = m->bridge->ofproto;
5043 uint64_t tx_packets, tx_bytes;
5044 const char *keys[2];
5046 size_t stat_cnt = 0;
5048 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5049 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5053 if (tx_packets != UINT64_MAX) {
5054 keys[stat_cnt] = "tx_packets";
5055 values[stat_cnt] = tx_packets;
5058 if (tx_bytes != UINT64_MAX) {
5059 keys[stat_cnt] = "tx_bytes";
5060 values[stat_cnt] = tx_bytes;
5064 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5068 * Add registered netdev and dpif types to ovsdb to allow external
5069 * applications to query the capabilities of the Open vSwitch instance
5070 * running on the node.
5073 discover_types(const struct ovsrec_open_vswitch *cfg)
5077 /* Datapath types. */
5079 dp_enumerate_types(&types);
5080 const char **datapath_types = sset_array(&types);
5081 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5082 sset_count(&types));
5083 free(datapath_types);
5084 sset_destroy(&types);
5088 netdev_enumerate_types(&types);
5089 const char **iface_types = sset_array(&types);
5090 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5092 sset_destroy(&types);