1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
51 #include "if-notifier.h"
56 #include "socket-util.h"
58 #include "stream-ssl.h"
60 #include "system-stats.h"
65 #include "lib/vswitch-idl.h"
66 #include "xenserver.h"
67 #include "openvswitch/vlog.h"
68 #include "sflow_api.h"
69 #include "vlan-bitmap.h"
72 VLOG_DEFINE_THIS_MODULE(bridge);
74 COVERAGE_DEFINE(bridge_reconfigure);
77 /* These members are always valid.
79 * They are immutable: they never change between iface_create() and
81 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
82 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
83 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
84 struct port *port; /* Containing port. */
85 char *name; /* Host network device name. */
86 struct netdev *netdev; /* Network device. */
87 ofp_port_t ofp_port; /* OpenFlow port number. */
90 /* These members are valid only within bridge_reconfigure(). */
91 const char *type; /* Usually same as cfg->type. */
92 const struct ovsrec_interface *cfg;
96 struct uuid uuid; /* UUID of this "mirror" record in database. */
97 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
98 struct bridge *bridge;
100 const struct ovsrec_mirror *cfg;
104 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
105 struct bridge *bridge;
108 const struct ovsrec_port *cfg;
110 /* An ordinary bridge port has 1 interface.
111 * A bridge port for bonding has at least 2 interfaces. */
112 struct ovs_list ifaces; /* List of "struct iface"s. */
116 struct hmap_node node; /* In 'all_bridges'. */
117 char *name; /* User-specified arbitrary name. */
118 char *type; /* Datapath type. */
119 struct eth_addr ea; /* Bridge Ethernet Address. */
120 struct eth_addr default_ea; /* Default MAC. */
121 const struct ovsrec_bridge *cfg;
123 /* OpenFlow switch processing. */
124 struct ofproto *ofproto; /* OpenFlow switch. */
127 struct hmap ports; /* "struct port"s indexed by name. */
128 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
129 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
131 /* Port mirroring. */
132 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
135 struct hmap mappings; /* "struct" indexed by UUID */
137 /* Used during reconfiguration. */
138 struct shash wanted_ports;
140 /* Synthetic local port if necessary. */
141 struct ovsrec_port synth_local_port;
142 struct ovsrec_interface synth_local_iface;
143 struct ovsrec_interface *synth_local_ifacep;
147 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
148 struct bridge *bridge;
154 /* All bridges, indexed by name. */
155 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
157 /* OVSDB IDL used to obtain configuration. */
158 static struct ovsdb_idl *idl;
160 /* We want to complete daemonization, fully detaching from our parent process,
161 * only after we have completed our initial configuration, committed our state
162 * to the database, and received confirmation back from the database server
163 * that it applied the commit. This allows our parent process to know that,
164 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
165 * to have already been filled in. (It is only "very likely" rather than
166 * certain because there is always a slim possibility that the transaction will
167 * fail or that some other client has added new bridges, ports, etc. while
168 * ovs-vswitchd was configuring using an old configuration.)
170 * We only need to do this once for our initial configuration at startup, so
171 * 'initial_config_done' tracks whether we've already done it. While we are
172 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
173 * itself and is otherwise NULL. */
174 static bool initial_config_done;
175 static struct ovsdb_idl_txn *daemonize_txn;
177 /* Most recently processed IDL sequence number. */
178 static unsigned int idl_seqno;
180 /* Track changes to port connectivity. */
181 static uint64_t connectivity_seqno = LLONG_MIN;
183 /* Status update to database.
185 * Some information in the database must be kept as up-to-date as possible to
186 * allow controllers to respond rapidly to network outages. Those status are
187 * updated via the 'status_txn'.
189 * We use the global connectivity sequence number to detect the status change.
190 * Also, to prevent the status update from sending too much to the database,
191 * we check the return status of each update transaction and do not start new
192 * update if the previous transaction status is 'TXN_INCOMPLETE'.
194 * 'statux_txn' is NULL if there is no ongoing status update.
196 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
197 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
198 * which will cause the main thread wake up soon and retry the status update.
200 static struct ovsdb_idl_txn *status_txn;
201 static bool status_txn_try_again;
203 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
204 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
205 #define STATUS_CHECK_AGAIN_MSEC 100
207 /* Statistics update to database. */
208 static struct ovsdb_idl_txn *stats_txn;
210 /* Each time this timer expires, the bridge fetches interface and mirror
211 * statistics and pushes them into the database. */
212 static int stats_timer_interval;
213 static long long int stats_timer = LLONG_MIN;
215 /* Each time this timer expires, the bridge fetches the list of port/VLAN
216 * membership that has been modified by the AA.
218 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
219 static long long int aa_refresh_timer = LLONG_MIN;
221 /* Whenever system interfaces are added, removed or change state, the bridge
222 * will be reconfigured.
224 static struct if_notifier *ifnotifier;
225 static bool ifaces_changed = false;
227 static void add_del_bridges(const struct ovsrec_open_vswitch *);
228 static void bridge_run__(void);
229 static void bridge_create(const struct ovsrec_bridge *);
230 static void bridge_destroy(struct bridge *);
231 static struct bridge *bridge_lookup(const char *name);
232 static unixctl_cb_func bridge_unixctl_dump_flows;
233 static unixctl_cb_func bridge_unixctl_reconnect;
234 static size_t bridge_get_controllers(const struct bridge *br,
235 struct ovsrec_controller ***controllersp);
236 static void bridge_collect_wanted_ports(struct bridge *,
237 const unsigned long *splinter_vlans,
238 struct shash *wanted_ports);
239 static void bridge_delete_ofprotos(void);
240 static void bridge_delete_or_reconfigure_ports(struct bridge *);
241 static void bridge_del_ports(struct bridge *,
242 const struct shash *wanted_ports);
243 static void bridge_add_ports(struct bridge *,
244 const struct shash *wanted_ports);
246 static void bridge_configure_datapath_id(struct bridge *);
247 static void bridge_configure_netflow(struct bridge *);
248 static void bridge_configure_forward_bpdu(struct bridge *);
249 static void bridge_configure_mac_table(struct bridge *);
250 static void bridge_configure_mcast_snooping(struct bridge *);
251 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
252 static void bridge_configure_ipfix(struct bridge *);
253 static void bridge_configure_spanning_tree(struct bridge *);
254 static void bridge_configure_tables(struct bridge *);
255 static void bridge_configure_dp_desc(struct bridge *);
256 static void bridge_configure_aa(struct bridge *);
257 static void bridge_aa_refresh_queued(struct bridge *);
258 static bool bridge_aa_need_refresh(struct bridge *);
259 static void bridge_configure_remotes(struct bridge *,
260 const struct sockaddr_in *managers,
262 static void bridge_pick_local_hw_addr(struct bridge *, struct eth_addr *ea,
263 struct iface **hw_addr_iface);
264 static uint64_t bridge_pick_datapath_id(struct bridge *,
265 const struct eth_addr bridge_ea,
266 struct iface *hw_addr_iface);
267 static uint64_t dpid_from_hash(const void *, size_t nbytes);
268 static bool bridge_has_bond_fake_iface(const struct bridge *,
270 static bool port_is_bond_fake_iface(const struct port *);
272 static unixctl_cb_func qos_unixctl_show;
274 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
275 static void port_del_ifaces(struct port *);
276 static void port_destroy(struct port *);
277 static struct port *port_lookup(const struct bridge *, const char *name);
278 static void port_configure(struct port *);
279 static struct lacp_settings *port_configure_lacp(struct port *,
280 struct lacp_settings *);
281 static void port_configure_bond(struct port *, struct bond_settings *);
282 static bool port_is_synthetic(const struct port *);
284 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
285 static void run_system_stats(void);
287 static void bridge_configure_mirrors(struct bridge *);
288 static struct mirror *mirror_create(struct bridge *,
289 const struct ovsrec_mirror *);
290 static void mirror_destroy(struct mirror *);
291 static bool mirror_configure(struct mirror *);
292 static void mirror_refresh_stats(struct mirror *);
294 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
295 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
296 const struct ovsrec_port *);
297 static bool iface_is_internal(const struct ovsrec_interface *iface,
298 const struct ovsrec_bridge *br);
299 static const char *iface_get_type(const struct ovsrec_interface *,
300 const struct ovsrec_bridge *);
301 static void iface_destroy(struct iface *);
302 static void iface_destroy__(struct iface *);
303 static struct iface *iface_lookup(const struct bridge *, const char *name);
304 static struct iface *iface_find(const char *name);
305 static struct iface *iface_from_ofp_port(const struct bridge *,
306 ofp_port_t ofp_port);
307 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
308 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
309 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
310 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
311 static void iface_configure_cfm(struct iface *);
312 static void iface_refresh_cfm_stats(struct iface *);
313 static void iface_refresh_stats(struct iface *);
314 static void iface_refresh_netdev_status(struct iface *);
315 static void iface_refresh_ofproto_status(struct iface *);
316 static bool iface_is_synthetic(const struct iface *);
317 static ofp_port_t iface_get_requested_ofp_port(
318 const struct ovsrec_interface *);
319 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
322 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
324 * This is deprecated. It is only for compatibility with broken device drivers
325 * in old versions of Linux that do not properly support VLANs when VLAN
326 * devices are not used. When broken device drivers are no longer in
327 * widespread use, we will delete these interfaces. */
329 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
330 static bool vlan_splinters_enabled_anywhere;
332 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
333 static unsigned long int *collect_splinter_vlans(
334 const struct ovsrec_open_vswitch *);
335 static void configure_splinter_port(struct port *);
336 static void add_vlan_splinter_ports(struct bridge *,
337 const unsigned long int *splinter_vlans,
338 struct shash *ports);
340 static void discover_types(const struct ovsrec_open_vswitch *cfg);
343 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
345 struct shash iface_hints;
346 static bool initialized = false;
353 shash_init(&iface_hints);
356 for (i = 0; i < cfg->n_bridges; i++) {
357 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
360 for (j = 0; j < br_cfg->n_ports; j++) {
361 struct ovsrec_port *port_cfg = br_cfg->ports[j];
364 for (k = 0; k < port_cfg->n_interfaces; k++) {
365 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
366 struct iface_hint *iface_hint;
368 iface_hint = xmalloc(sizeof *iface_hint);
369 iface_hint->br_name = br_cfg->name;
370 iface_hint->br_type = br_cfg->datapath_type;
371 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
373 shash_add(&iface_hints, if_cfg->name, iface_hint);
379 ofproto_init(&iface_hints);
381 shash_destroy_free_data(&iface_hints);
386 if_change_cb(void *aux OVS_UNUSED)
388 ifaces_changed = true;
391 /* Public functions. */
393 /* Initializes the bridge module, configuring it to obtain its configuration
394 * from an OVSDB server accessed over 'remote', which should be a string in a
395 * form acceptable to ovsdb_idl_create(). */
397 bridge_init(const char *remote)
399 /* Create connection to database. */
400 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
401 idl_seqno = ovsdb_idl_get_seqno(idl);
402 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
403 ovsdb_idl_verify_write_only(idl);
405 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
406 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
407 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
408 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
409 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
410 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
411 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
412 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
413 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
415 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
416 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
417 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
418 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
419 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
420 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
421 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
423 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
426 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
427 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
428 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
429 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
430 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
446 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
447 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
448 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
449 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
450 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
451 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
453 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
454 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
455 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
456 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
458 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
460 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
462 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
463 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
465 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
466 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
467 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
468 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
470 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
471 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
472 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
473 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
474 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
476 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
478 /* Register unixctl commands. */
479 unixctl_command_register("qos/show", "interface", 1, 1,
480 qos_unixctl_show, NULL);
481 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
482 bridge_unixctl_dump_flows, NULL);
483 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
484 bridge_unixctl_reconnect, NULL);
493 ifnotifier = if_notifier_create(if_change_cb, NULL);
499 struct bridge *br, *next_br;
501 if_notifier_destroy(ifnotifier);
502 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
505 ovsdb_idl_destroy(idl);
508 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
509 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
510 * responsible for freeing '*managersp' (with free()).
512 * You may be asking yourself "why does ovs-vswitchd care?", because
513 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
514 * should not be and in fact is not directly involved in that. But
515 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
516 * it has to tell in-band control where the managers are to enable that.
517 * (Thus, only managers connected in-band and with non-loopback addresses
521 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
522 struct sockaddr_in **managersp, size_t *n_managersp)
524 struct sockaddr_in *managers = NULL;
525 size_t n_managers = 0;
529 /* Collect all of the potential targets from the "targets" columns of the
530 * rows pointed to by "manager_options", excluding any that are
533 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
534 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
536 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
537 sset_find_and_delete(&targets, m->target);
539 sset_add(&targets, m->target);
543 /* Now extract the targets' IP addresses. */
544 if (!sset_is_empty(&targets)) {
547 managers = xmalloc(sset_count(&targets) * sizeof *managers);
548 SSET_FOR_EACH (target, &targets) {
550 struct sockaddr_storage ss;
551 struct sockaddr_in in;
554 /* Ignore loopback. */
555 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
557 && sa.ss.ss_family == AF_INET
558 && sa.in.sin_addr.s_addr != htonl(INADDR_LOOPBACK)) {
559 managers[n_managers++] = sa.in;
563 sset_destroy(&targets);
565 *managersp = managers;
566 *n_managersp = n_managers;
570 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
572 unsigned long int *splinter_vlans;
573 struct sockaddr_in *managers;
574 struct bridge *br, *next;
575 int sflow_bridge_number;
578 COVERAGE_INC(bridge_reconfigure);
580 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
581 OFPROTO_FLOW_LIMIT_DEFAULT));
582 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
583 OFPROTO_MAX_IDLE_DEFAULT));
584 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
586 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
589 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
590 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
592 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
593 * to 'ovs_cfg', with only very minimal configuration otherwise.
595 * This is mostly an update to bridge data structures. Nothing is pushed
596 * down to ofproto or lower layers. */
597 add_del_bridges(ovs_cfg);
598 splinter_vlans = collect_splinter_vlans(ovs_cfg);
599 HMAP_FOR_EACH (br, node, &all_bridges) {
600 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
601 bridge_del_ports(br, &br->wanted_ports);
603 free(splinter_vlans);
605 /* Start pushing configuration changes down to the ofproto layer:
607 * - Delete ofprotos that are no longer configured.
609 * - Delete ports that are no longer configured.
611 * - Reconfigure existing ports to their desired configurations, or
612 * delete them if not possible.
614 * We have to do all the deletions before we can do any additions, because
615 * the ports to be added might require resources that will be freed up by
616 * deletions (they might especially overlap in name). */
617 bridge_delete_ofprotos();
618 HMAP_FOR_EACH (br, node, &all_bridges) {
620 bridge_delete_or_reconfigure_ports(br);
624 /* Finish pushing configuration changes to the ofproto layer:
626 * - Create ofprotos that are missing.
628 * - Add ports that are missing. */
629 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
633 error = ofproto_create(br->name, br->type, &br->ofproto);
635 VLOG_ERR("failed to create bridge %s: %s", br->name,
636 ovs_strerror(error));
637 shash_destroy(&br->wanted_ports);
640 /* Trigger storing datapath version. */
641 seq_change(connectivity_seq_get());
645 HMAP_FOR_EACH (br, node, &all_bridges) {
646 bridge_add_ports(br, &br->wanted_ports);
647 shash_destroy(&br->wanted_ports);
650 reconfigure_system_stats(ovs_cfg);
652 /* Complete the configuration. */
653 sflow_bridge_number = 0;
654 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
655 HMAP_FOR_EACH (br, node, &all_bridges) {
658 /* We need the datapath ID early to allow LACP ports to use it as the
659 * default system ID. */
660 bridge_configure_datapath_id(br);
662 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
665 port_configure(port);
667 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
668 iface_set_ofport(iface->cfg, iface->ofp_port);
669 /* Clear eventual previous errors */
670 ovsrec_interface_set_error(iface->cfg, NULL);
671 iface_configure_cfm(iface);
672 iface_configure_qos(iface, port->cfg->qos);
673 iface_set_mac(br, port, iface);
674 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
676 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
680 bridge_configure_mirrors(br);
681 bridge_configure_forward_bpdu(br);
682 bridge_configure_mac_table(br);
683 bridge_configure_mcast_snooping(br);
684 bridge_configure_remotes(br, managers, n_managers);
685 bridge_configure_netflow(br);
686 bridge_configure_sflow(br, &sflow_bridge_number);
687 bridge_configure_ipfix(br);
688 bridge_configure_spanning_tree(br);
689 bridge_configure_tables(br);
690 bridge_configure_dp_desc(br);
691 bridge_configure_aa(br);
695 /* The ofproto-dpif provider does some final reconfiguration in its
696 * ->type_run() function. We have to call it before notifying the database
697 * client that reconfiguration is complete, otherwise there is a very
698 * narrow race window in which e.g. ofproto/trace will not recognize the
699 * new configuration (sometimes this causes unit test failures). */
703 /* Delete ofprotos which aren't configured or have the wrong type. Create
704 * ofprotos which don't exist but need to. */
706 bridge_delete_ofprotos(void)
713 /* Delete ofprotos with no bridge or with the wrong type. */
716 ofproto_enumerate_types(&types);
717 SSET_FOR_EACH (type, &types) {
720 ofproto_enumerate_names(type, &names);
721 SSET_FOR_EACH (name, &names) {
722 br = bridge_lookup(name);
723 if (!br || strcmp(type, br->type)) {
724 ofproto_delete(name, type);
728 sset_destroy(&names);
729 sset_destroy(&types);
733 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
735 if (*n >= *allocated) {
736 ports = x2nrealloc(ports, allocated, sizeof *ports);
738 ports[(*n)++] = port;
743 bridge_delete_or_reconfigure_ports(struct bridge *br)
745 struct ofproto_port ofproto_port;
746 struct ofproto_port_dump dump;
748 struct sset ofproto_ports;
749 struct port *port, *port_next;
751 /* List of "ofp_port"s to delete. We make a list instead of deleting them
752 * right away because ofproto implementations aren't necessarily able to
753 * iterate through a changing list of ports in an entirely robust way. */
760 sset_init(&ofproto_ports);
762 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
763 * that are not configured in the database. (This commonly happens when
764 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
766 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
767 * that have the wrong OpenFlow port number (and arrange to add them back
768 * with the correct OpenFlow port number). */
769 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
770 ofp_port_t requested_ofp_port;
773 sset_add(&ofproto_ports, ofproto_port.name);
775 iface = iface_lookup(br, ofproto_port.name);
777 /* No such iface is configured, so we should delete this
780 * As a corner case exception, keep the port if it's a bond fake
782 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
783 && !strcmp(ofproto_port.type, "internal")) {
789 if (strcmp(ofproto_port.type, iface->type)
790 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
791 /* The interface is the wrong type or can't be configured.
796 /* If the requested OpenFlow port for 'iface' changed, and it's not
797 * already the correct port, then we might want to temporarily delete
798 * this interface, so we can add it back again with the new OpenFlow
800 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
801 if (iface->ofp_port != OFPP_LOCAL &&
802 requested_ofp_port != OFPP_NONE &&
803 requested_ofp_port != iface->ofp_port) {
804 ofp_port_t victim_request;
805 struct iface *victim;
807 /* Check for an existing OpenFlow port currently occupying
808 * 'iface''s requested port number. If there isn't one, then
809 * delete this port. Otherwise we need to consider further. */
810 victim = iface_from_ofp_port(br, requested_ofp_port);
815 /* 'victim' is a port currently using 'iface''s requested port
816 * number. Unless 'victim' specifically requested that port
817 * number, too, then we can delete both 'iface' and 'victim'
818 * temporarily. (We'll add both of them back again later with new
819 * OpenFlow port numbers.)
821 * If 'victim' did request port number 'requested_ofp_port', just
822 * like 'iface', then that's a configuration inconsistency that we
823 * can't resolve. We might as well let it keep its current port
825 victim_request = iface_get_requested_ofp_port(victim->cfg);
826 if (victim_request != requested_ofp_port) {
827 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
828 iface_destroy(victim);
837 iface_destroy(iface);
838 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
840 for (i = 0; i < n; i++) {
841 ofproto_port_del(br->ofproto, del[i]);
845 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
846 * that we didn't see when we iterated through the datapath, i.e. ports
847 * that disappeared underneath use. This is an unusual situation, but it
848 * can happen in some cases:
850 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
851 * idea but could happen).
853 * - The port represented a device that disappeared, e.g. a tuntap
854 * device destroyed via "tunctl -d", a physical Ethernet device
855 * whose module was just unloaded via "rmmod", or a virtual NIC for a
856 * VM whose VM was just terminated. */
857 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
858 struct iface *iface, *iface_next;
860 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
861 if (!sset_contains(&ofproto_ports, iface->name)) {
862 iface_destroy__(iface);
866 if (list_is_empty(&port->ifaces)) {
870 sset_destroy(&ofproto_ports);
874 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
875 bool with_requested_port)
877 struct shash_node *port_node;
879 SHASH_FOR_EACH (port_node, wanted_ports) {
880 const struct ovsrec_port *port_cfg = port_node->data;
883 for (i = 0; i < port_cfg->n_interfaces; i++) {
884 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
885 ofp_port_t requested_ofp_port;
887 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
888 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
889 struct iface *iface = iface_lookup(br, iface_cfg->name);
892 iface_create(br, iface_cfg, port_cfg);
900 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
902 /* First add interfaces that request a particular port number. */
903 bridge_add_ports__(br, wanted_ports, true);
905 /* Then add interfaces that want automatic port number assignment.
906 * We add these afterward to avoid accidentally taking a specifically
907 * requested port number. */
908 bridge_add_ports__(br, wanted_ports, false);
912 port_configure(struct port *port)
914 const struct ovsrec_port *cfg = port->cfg;
915 struct bond_settings bond_settings;
916 struct lacp_settings lacp_settings;
917 struct ofproto_bundle_settings s;
920 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
921 configure_splinter_port(port);
930 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
931 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
932 s.slaves[s.n_slaves++] = iface->ofp_port;
937 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
941 /* Get VLAN trunks. */
944 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
948 if (cfg->vlan_mode) {
949 if (!strcmp(cfg->vlan_mode, "access")) {
950 s.vlan_mode = PORT_VLAN_ACCESS;
951 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
952 s.vlan_mode = PORT_VLAN_TRUNK;
953 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
954 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
955 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
956 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
958 /* This "can't happen" because ovsdb-server should prevent it. */
959 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
960 "back to trunk mode", port->name, cfg->vlan_mode);
961 s.vlan_mode = PORT_VLAN_TRUNK;
965 s.vlan_mode = PORT_VLAN_ACCESS;
967 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
971 s.vlan_mode = PORT_VLAN_TRUNK;
974 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
977 /* Get LACP settings. */
978 s.lacp = port_configure_lacp(port, &lacp_settings);
982 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
983 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
984 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
987 s.lacp_slaves = NULL;
990 /* Get bond settings. */
991 if (s.n_slaves > 1) {
992 s.bond = &bond_settings;
993 port_configure_bond(port, &bond_settings);
996 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
997 netdev_set_miimon_interval(iface->netdev, 0);
1002 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1007 free(s.lacp_slaves);
1010 /* Pick local port hardware address and datapath ID for 'br'. */
1012 bridge_configure_datapath_id(struct bridge *br)
1016 struct iface *local_iface;
1017 struct iface *hw_addr_iface;
1020 bridge_pick_local_hw_addr(br, &ea, &hw_addr_iface);
1021 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1023 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1025 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1026 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1027 "Ethernet address: %s",
1028 br->name, ovs_strerror(error));
1033 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1034 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1035 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1036 ofproto_set_datapath_id(br->ofproto, dpid);
1039 dpid_string = xasprintf("%016"PRIx64, dpid);
1040 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1044 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1047 bridge_get_allowed_versions(struct bridge *br)
1049 if (!br->cfg->n_protocols) {
1053 return ofputil_versions_from_strings(br->cfg->protocols,
1054 br->cfg->n_protocols);
1057 /* Set NetFlow configuration on 'br'. */
1059 bridge_configure_netflow(struct bridge *br)
1061 struct ovsrec_netflow *cfg = br->cfg->netflow;
1062 struct netflow_options opts;
1065 ofproto_set_netflow(br->ofproto, NULL);
1069 memset(&opts, 0, sizeof opts);
1071 /* Get default NetFlow configuration from datapath.
1072 * Apply overrides from 'cfg'. */
1073 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1074 if (cfg->engine_type) {
1075 opts.engine_type = *cfg->engine_type;
1077 if (cfg->engine_id) {
1078 opts.engine_id = *cfg->engine_id;
1081 /* Configure active timeout interval. */
1082 opts.active_timeout = cfg->active_timeout;
1083 if (!opts.active_timeout) {
1084 opts.active_timeout = -1;
1085 } else if (opts.active_timeout < 0) {
1086 VLOG_WARN("bridge %s: active timeout interval set to negative "
1087 "value, using default instead (%d seconds)", br->name,
1088 NF_ACTIVE_TIMEOUT_DEFAULT);
1089 opts.active_timeout = -1;
1092 /* Add engine ID to interface number to disambiguate bridgs? */
1093 opts.add_id_to_iface = cfg->add_id_to_interface;
1094 if (opts.add_id_to_iface) {
1095 if (opts.engine_id > 0x7f) {
1096 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1097 "another vswitch, choose an engine id less than 128",
1100 if (hmap_count(&br->ports) > 508) {
1101 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1102 "another port when more than 508 ports are used",
1108 sset_init(&opts.collectors);
1109 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1112 if (ofproto_set_netflow(br->ofproto, &opts)) {
1113 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1115 sset_destroy(&opts.collectors);
1118 /* Set sFlow configuration on 'br'. */
1120 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1122 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1123 struct ovsrec_controller **controllers;
1124 struct ofproto_sflow_options oso;
1125 size_t n_controllers;
1129 ofproto_set_sflow(br->ofproto, NULL);
1133 memset(&oso, 0, sizeof oso);
1135 sset_init(&oso.targets);
1136 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1138 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1139 if (cfg->sampling) {
1140 oso.sampling_rate = *cfg->sampling;
1143 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1145 oso.polling_interval = *cfg->polling;
1148 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1150 oso.header_len = *cfg->header;
1153 oso.sub_id = (*sflow_bridge_number)++;
1154 oso.agent_device = cfg->agent;
1156 oso.control_ip = NULL;
1157 n_controllers = bridge_get_controllers(br, &controllers);
1158 for (i = 0; i < n_controllers; i++) {
1159 if (controllers[i]->local_ip) {
1160 oso.control_ip = controllers[i]->local_ip;
1164 ofproto_set_sflow(br->ofproto, &oso);
1166 sset_destroy(&oso.targets);
1169 /* Returns whether a IPFIX row is valid. */
1171 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1173 return ipfix && ipfix->n_targets > 0;
1176 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1178 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1179 const struct bridge *br)
1181 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1184 /* Set IPFIX configuration on 'br'. */
1186 bridge_configure_ipfix(struct bridge *br)
1188 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1189 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1190 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1191 struct ofproto_ipfix_bridge_exporter_options be_opts;
1192 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1193 size_t n_fe_opts = 0;
1195 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1196 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1201 if (!valid_be_cfg && n_fe_opts == 0) {
1202 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1207 memset(&be_opts, 0, sizeof be_opts);
1209 sset_init(&be_opts.targets);
1210 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1212 if (be_cfg->sampling) {
1213 be_opts.sampling_rate = *be_cfg->sampling;
1215 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1217 if (be_cfg->obs_domain_id) {
1218 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1220 if (be_cfg->obs_point_id) {
1221 be_opts.obs_point_id = *be_cfg->obs_point_id;
1223 if (be_cfg->cache_active_timeout) {
1224 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1226 if (be_cfg->cache_max_flows) {
1227 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1230 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1231 "enable-tunnel-sampling", true);
1233 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1234 "enable-input-sampling", false);
1236 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1237 "enable-output-sampling", false);
1240 if (n_fe_opts > 0) {
1241 struct ofproto_ipfix_flow_exporter_options *opts;
1242 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1244 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1245 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1246 opts->collector_set_id = fe_cfg->id;
1247 sset_init(&opts->targets);
1248 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1249 fe_cfg->ipfix->n_targets);
1250 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1251 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1252 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1253 ? *fe_cfg->ipfix->cache_max_flows : 0;
1259 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1263 sset_destroy(&be_opts.targets);
1266 if (n_fe_opts > 0) {
1267 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1269 for (i = 0; i < n_fe_opts; i++) {
1270 sset_destroy(&opts->targets);
1278 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1279 struct ofproto_port_stp_settings *port_s,
1280 int *port_num_counter, unsigned long *port_num_bitmap)
1282 const char *config_str;
1283 struct iface *iface;
1285 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1286 port_s->enable = false;
1289 port_s->enable = true;
1292 /* STP over bonds is not supported. */
1293 if (!list_is_singleton(&port->ifaces)) {
1294 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1296 port_s->enable = false;
1300 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1302 /* Internal ports shouldn't participate in spanning tree, so
1304 if (!strcmp(iface->type, "internal")) {
1305 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1306 port_s->enable = false;
1310 /* STP on mirror output ports is not supported. */
1311 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1312 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1313 port_s->enable = false;
1317 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1319 unsigned long int port_num = strtoul(config_str, NULL, 0);
1320 int port_idx = port_num - 1;
1322 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1323 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1324 port_s->enable = false;
1328 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1329 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1330 port->name, port_num);
1331 port_s->enable = false;
1334 bitmap_set1(port_num_bitmap, port_idx);
1335 port_s->port_num = port_idx;
1337 if (*port_num_counter >= STP_MAX_PORTS) {
1338 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1339 port_s->enable = false;
1343 port_s->port_num = (*port_num_counter)++;
1346 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1348 port_s->path_cost = strtoul(config_str, NULL, 10);
1350 enum netdev_features current;
1353 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1354 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1355 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1358 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1360 port_s->priority = strtoul(config_str, NULL, 0);
1362 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1367 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1368 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1370 const char *config_str;
1371 struct iface *iface;
1373 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1374 port_s->enable = false;
1377 port_s->enable = true;
1380 /* RSTP over bonds is not supported. */
1381 if (!list_is_singleton(&port->ifaces)) {
1382 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1384 port_s->enable = false;
1388 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1390 /* Internal ports shouldn't participate in spanning tree, so
1392 if (!strcmp(iface->type, "internal")) {
1393 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1394 port_s->enable = false;
1398 /* RSTP on mirror output ports is not supported. */
1399 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1400 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1401 port_s->enable = false;
1405 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1407 unsigned long int port_num = strtoul(config_str, NULL, 0);
1408 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1409 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1410 port_s->enable = false;
1413 port_s->port_num = port_num;
1415 if (*port_num_counter >= RSTP_MAX_PORTS) {
1416 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1417 port_s->enable = false;
1420 /* If rstp-port-num is not specified, use 0.
1421 * rstp_port_set_port_number() will look for the first free one. */
1422 port_s->port_num = 0;
1425 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1427 port_s->path_cost = strtoul(config_str, NULL, 10);
1429 enum netdev_features current;
1432 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1433 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1434 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1437 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1439 port_s->priority = strtoul(config_str, NULL, 0);
1441 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1444 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1446 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1448 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1451 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1452 "rstp-admin-port-state", true);
1454 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1455 "rstp-port-admin-edge", false);
1456 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1457 "rstp-port-auto-edge", true);
1458 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1459 "rstp-port-mcheck", false);
1462 /* Set spanning tree configuration on 'br'. */
1464 bridge_configure_stp(struct bridge *br, bool enable_stp)
1467 ofproto_set_stp(br->ofproto, NULL);
1469 struct ofproto_stp_settings br_s;
1470 const char *config_str;
1472 int port_num_counter;
1473 unsigned long *port_num_bitmap;
1475 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1479 if (eth_addr_from_string(config_str, &ea)) {
1480 br_s.system_id = eth_addr_to_uint64(ea);
1482 br_s.system_id = eth_addr_to_uint64(br->ea);
1483 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1484 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1487 br_s.system_id = eth_addr_to_uint64(br->ea);
1490 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1492 br_s.priority = strtoul(config_str, NULL, 0);
1494 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1497 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1499 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1501 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1504 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1506 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1508 br_s.max_age = STP_DEFAULT_MAX_AGE;
1511 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1513 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1515 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1518 /* Configure STP on the bridge. */
1519 if (ofproto_set_stp(br->ofproto, &br_s)) {
1520 VLOG_ERR("bridge %s: could not enable STP", br->name);
1524 /* Users must either set the port number with the "stp-port-num"
1525 * configuration on all ports or none. If manual configuration
1526 * is not done, then we allocate them sequentially. */
1527 port_num_counter = 0;
1528 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1529 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1530 struct ofproto_port_stp_settings port_s;
1531 struct iface *iface;
1533 port_configure_stp(br->ofproto, port, &port_s,
1534 &port_num_counter, port_num_bitmap);
1536 /* As bonds are not supported, just apply configuration to
1537 * all interfaces. */
1538 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1539 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1541 VLOG_ERR("port %s: could not enable STP", port->name);
1547 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1548 && port_num_counter) {
1549 VLOG_ERR("bridge %s: must manually configure all STP port "
1550 "IDs or none, disabling", br->name);
1551 ofproto_set_stp(br->ofproto, NULL);
1553 bitmap_free(port_num_bitmap);
1558 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1561 ofproto_set_rstp(br->ofproto, NULL);
1563 struct ofproto_rstp_settings br_s;
1564 const char *config_str;
1566 int port_num_counter;
1568 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1572 if (eth_addr_from_string(config_str, &ea)) {
1573 br_s.address = eth_addr_to_uint64(ea);
1576 br_s.address = eth_addr_to_uint64(br->ea);
1577 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1578 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1582 br_s.address = eth_addr_to_uint64(br->ea);
1585 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1587 br_s.priority = strtoul(config_str, NULL, 0);
1589 br_s.priority = RSTP_DEFAULT_PRIORITY;
1592 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1594 br_s.ageing_time = strtoul(config_str, NULL, 0);
1596 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1599 config_str = smap_get(&br->cfg->other_config,
1600 "rstp-force-protocol-version");
1602 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1604 br_s.force_protocol_version = FPV_DEFAULT;
1607 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1609 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1611 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1614 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1616 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1618 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1621 config_str = smap_get(&br->cfg->other_config,
1622 "rstp-transmit-hold-count");
1624 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1626 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1629 /* Configure RSTP on the bridge. */
1630 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1631 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1635 port_num_counter = 0;
1636 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1637 struct ofproto_port_rstp_settings port_s;
1638 struct iface *iface;
1640 port_configure_rstp(br->ofproto, port, &port_s,
1643 /* As bonds are not supported, just apply configuration to
1644 * all interfaces. */
1645 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1646 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1648 VLOG_ERR("port %s: could not enable RSTP", port->name);
1657 bridge_configure_spanning_tree(struct bridge *br)
1659 bool enable_rstp = br->cfg->rstp_enable;
1660 bool enable_stp = br->cfg->stp_enable;
1662 if (enable_rstp && enable_stp) {
1663 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1664 "configured; enabling RSTP", br->name);
1668 bridge_configure_stp(br, enable_stp);
1669 bridge_configure_rstp(br, enable_rstp);
1673 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1675 const struct port *port = port_lookup(br, name);
1676 return port && port_is_bond_fake_iface(port);
1680 port_is_bond_fake_iface(const struct port *port)
1682 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1686 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1688 struct bridge *br, *next;
1689 struct shash_node *node;
1690 struct shash new_br;
1693 /* Collect new bridges' names and types. */
1694 shash_init(&new_br);
1695 for (i = 0; i < cfg->n_bridges; i++) {
1696 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1697 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1699 if (strchr(br_cfg->name, '/') || strchr(br_cfg->name, '\\')) {
1700 /* Prevent remote ovsdb-server users from accessing arbitrary
1701 * directories, e.g. consider a bridge named "../../../etc/".
1703 * Prohibiting "\" is only necessary on Windows but it's no great
1704 * loss elsewhere. */
1705 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1707 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1708 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1712 /* Get rid of deleted bridges or those whose types have changed.
1713 * Update 'cfg' of bridges that still exist. */
1714 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1715 br->cfg = shash_find_data(&new_br, br->name);
1716 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1717 br->cfg->datapath_type))) {
1722 /* Add new bridges. */
1723 SHASH_FOR_EACH(node, &new_br) {
1724 const struct ovsrec_bridge *br_cfg = node->data;
1725 struct bridge *br = bridge_lookup(br_cfg->name);
1727 bridge_create(br_cfg);
1731 shash_destroy(&new_br);
1734 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1735 * Returns 0 if successful, otherwise a positive errno value. */
1737 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1738 struct netdev *netdev, char **errp)
1740 return netdev_set_config(netdev, &iface_cfg->options, errp);
1743 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1744 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1746 * If successful, returns 0 and stores the network device in '*netdevp'. On
1747 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1749 iface_do_create(const struct bridge *br,
1750 const struct ovsrec_interface *iface_cfg,
1751 const struct ovsrec_port *port_cfg,
1752 ofp_port_t *ofp_portp, struct netdev **netdevp,
1755 struct netdev *netdev = NULL;
1758 if (netdev_is_reserved_name(iface_cfg->name)) {
1759 VLOG_WARN("could not create interface %s, name is reserved",
1765 error = netdev_open(iface_cfg->name,
1766 iface_get_type(iface_cfg, br->cfg), &netdev);
1768 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1769 iface_cfg->name, ovs_strerror(error));
1773 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1778 *ofp_portp = iface_pick_ofport(iface_cfg);
1779 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1784 VLOG_INFO("bridge %s: added interface %s on port %d",
1785 br->name, iface_cfg->name, *ofp_portp);
1787 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1788 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1796 netdev_close(netdev);
1800 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1801 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1802 * automatically allocated for the iface. Takes ownership of and
1803 * deallocates 'if_cfg'.
1805 * Return true if an iface is successfully created, false otherwise. */
1807 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1808 const struct ovsrec_port *port_cfg)
1810 struct netdev *netdev;
1811 struct iface *iface;
1812 ofp_port_t ofp_port;
1817 /* Do the bits that can fail up front. */
1818 ovs_assert(!iface_lookup(br, iface_cfg->name));
1819 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1821 iface_clear_db_record(iface_cfg, errp);
1826 /* Get or create the port structure. */
1827 port = port_lookup(br, port_cfg->name);
1829 port = port_create(br, port_cfg);
1832 /* Create the iface structure. */
1833 iface = xzalloc(sizeof *iface);
1834 list_push_back(&port->ifaces, &iface->port_elem);
1835 hmap_insert(&br->iface_by_name, &iface->name_node,
1836 hash_string(iface_cfg->name, 0));
1838 iface->name = xstrdup(iface_cfg->name);
1839 iface->ofp_port = ofp_port;
1840 iface->netdev = netdev;
1841 iface->type = iface_get_type(iface_cfg, br->cfg);
1842 iface->cfg = iface_cfg;
1843 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1844 hash_ofp_port(ofp_port));
1846 /* Populate initial status in database. */
1847 iface_refresh_stats(iface);
1848 iface_refresh_netdev_status(iface);
1850 /* Add bond fake iface if necessary. */
1851 if (port_is_bond_fake_iface(port)) {
1852 struct ofproto_port ofproto_port;
1854 if (ofproto_port_query_by_name(br->ofproto, port->name,
1856 struct netdev *netdev;
1859 error = netdev_open(port->name, "internal", &netdev);
1861 ofp_port_t fake_ofp_port = OFPP_NONE;
1862 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1863 netdev_close(netdev);
1865 VLOG_WARN("could not open network device %s (%s)",
1866 port->name, ovs_strerror(error));
1869 /* Already exists, nothing to do. */
1870 ofproto_port_destroy(&ofproto_port);
1877 /* Set forward BPDU option. */
1879 bridge_configure_forward_bpdu(struct bridge *br)
1881 ofproto_set_forward_bpdu(br->ofproto,
1882 smap_get_bool(&br->cfg->other_config,
1887 /* Set MAC learning table configuration for 'br'. */
1889 bridge_configure_mac_table(struct bridge *br)
1891 const char *idle_time_str;
1894 const char *mac_table_size_str;
1897 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1898 idle_time = (idle_time_str && atoi(idle_time_str)
1899 ? atoi(idle_time_str)
1900 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1902 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1903 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1904 ? atoi(mac_table_size_str)
1907 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1910 /* Set multicast snooping table configuration for 'br'. */
1912 bridge_configure_mcast_snooping(struct bridge *br)
1914 if (!br->cfg->mcast_snooping_enable) {
1915 ofproto_set_mcast_snooping(br->ofproto, NULL);
1918 struct ofproto_mcast_snooping_settings br_s;
1919 const char *idle_time_str;
1920 const char *max_entries_str;
1922 idle_time_str = smap_get(&br->cfg->other_config,
1923 "mcast-snooping-aging-time");
1924 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1925 ? atoi(idle_time_str)
1926 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1928 max_entries_str = smap_get(&br->cfg->other_config,
1929 "mcast-snooping-table-size");
1930 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1931 ? atoi(max_entries_str)
1932 : MCAST_DEFAULT_MAX_ENTRIES);
1934 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1935 "mcast-snooping-disable-flood-unregistered",
1938 /* Configure multicast snooping on the bridge */
1939 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1940 VLOG_ERR("bridge %s: could not enable multicast snooping",
1945 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1946 struct ofproto_mcast_snooping_port_settings port_s;
1947 port_s.flood = smap_get_bool(&port->cfg->other_config,
1948 "mcast-snooping-flood", false);
1949 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1950 "mcast-snooping-flood-reports", false);
1951 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1952 VLOG_ERR("port %s: could not configure mcast snooping",
1960 find_local_hw_addr(const struct bridge *br, struct eth_addr *ea,
1961 const struct port *fake_br, struct iface **hw_addr_iface)
1963 struct hmapx mirror_output_ports;
1965 bool found_addr = false;
1969 /* Mirror output ports don't participate in picking the local hardware
1970 * address. ofproto can't help us find out whether a given port is a
1971 * mirror output because we haven't configured mirrors yet, so we need to
1972 * accumulate them ourselves. */
1973 hmapx_init(&mirror_output_ports);
1974 for (i = 0; i < br->cfg->n_mirrors; i++) {
1975 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1976 if (m->output_port) {
1977 hmapx_add(&mirror_output_ports, m->output_port);
1981 /* Otherwise choose the minimum non-local MAC address among all of the
1983 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1984 struct eth_addr iface_ea;
1985 struct iface *candidate;
1986 struct iface *iface;
1988 /* Mirror output ports don't participate. */
1989 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1993 /* Choose the MAC address to represent the port. */
1995 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac,
1997 /* Find the interface with this Ethernet address (if any) so that
1998 * we can provide the correct devname to the caller. */
1999 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2000 struct eth_addr candidate_ea;
2001 if (!netdev_get_etheraddr(candidate->netdev, &candidate_ea)
2002 && eth_addr_equals(iface_ea, candidate_ea)) {
2007 /* Choose the interface whose MAC address will represent the port.
2008 * The Linux kernel bonding code always chooses the MAC address of
2009 * the first slave added to a bond, and the Fedora networking
2010 * scripts always add slaves to a bond in alphabetical order, so
2011 * for compatibility we choose the interface with the name that is
2012 * first in alphabetical order. */
2013 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2014 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2019 /* The local port doesn't count (since we're trying to choose its
2020 * MAC address anyway). */
2021 if (iface->ofp_port == OFPP_LOCAL) {
2025 /* For fake bridges we only choose from ports with the same tag */
2026 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2027 if (!port->cfg->tag) {
2030 if (*port->cfg->tag != *fake_br->cfg->tag) {
2036 error = netdev_get_etheraddr(iface->netdev, &iface_ea);
2042 /* Compare against our current choice. */
2043 if (!eth_addr_is_multicast(iface_ea) &&
2044 !eth_addr_is_local(iface_ea) &&
2045 !eth_addr_is_reserved(iface_ea) &&
2046 !eth_addr_is_zero(iface_ea) &&
2047 (!found_addr || eth_addr_compare_3way(iface_ea, *ea) < 0))
2050 *hw_addr_iface = iface;
2056 *ea = br->default_ea;
2057 *hw_addr_iface = NULL;
2060 hmapx_destroy(&mirror_output_ports);
2064 bridge_pick_local_hw_addr(struct bridge *br, struct eth_addr *ea,
2065 struct iface **hw_addr_iface)
2068 *hw_addr_iface = NULL;
2070 /* Did the user request a particular MAC? */
2071 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2072 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2073 if (eth_addr_is_multicast(*ea)) {
2074 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2075 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(*ea));
2076 } else if (eth_addr_is_zero(*ea)) {
2077 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2083 /* Find a local hw address */
2084 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2087 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2088 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2089 * an interface on 'br', then that interface must be passed in as
2090 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2091 * 'hw_addr_iface' must be passed in as a null pointer. */
2093 bridge_pick_datapath_id(struct bridge *br,
2094 const struct eth_addr bridge_ea,
2095 struct iface *hw_addr_iface)
2098 * The procedure for choosing a bridge MAC address will, in the most
2099 * ordinary case, also choose a unique MAC that we can use as a datapath
2100 * ID. In some special cases, though, multiple bridges will end up with
2101 * the same MAC address. This is OK for the bridges, but it will confuse
2102 * the OpenFlow controller, because each datapath needs a unique datapath
2105 * Datapath IDs must be unique. It is also very desirable that they be
2106 * stable from one run to the next, so that policy set on a datapath
2109 const char *datapath_id;
2112 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2113 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2117 if (!hw_addr_iface) {
2119 * A purely internal bridge, that is, one that has no non-virtual
2120 * network devices on it at all, is difficult because it has no
2121 * natural unique identifier at all.
2123 * When the host is a XenServer, we handle this case by hashing the
2124 * host's UUID with the name of the bridge. Names of bridges are
2125 * persistent across XenServer reboots, although they can be reused if
2126 * an internal network is destroyed and then a new one is later
2127 * created, so this is fairly effective.
2129 * When the host is not a XenServer, we punt by using a random MAC
2130 * address on each run.
2132 const char *host_uuid = xenserver_get_host_uuid();
2134 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2135 dpid = dpid_from_hash(combined, strlen(combined));
2141 return eth_addr_to_uint64(bridge_ea);
2145 dpid_from_hash(const void *data, size_t n)
2148 uint8_t bytes[SHA1_DIGEST_SIZE];
2152 sha1_bytes(data, n, hash.bytes);
2153 eth_addr_mark_random(&hash.ea);
2154 return eth_addr_to_uint64(hash.ea);
2158 iface_refresh_netdev_status(struct iface *iface)
2162 enum netdev_features current;
2163 enum netdev_flags flags;
2164 const char *link_state;
2165 struct eth_addr mac;
2166 int64_t bps, mtu_64, ifindex64, link_resets;
2169 if (iface_is_synthetic(iface)) {
2173 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2174 && !status_txn_try_again) {
2178 iface->change_seq = netdev_get_change_seq(iface->netdev);
2182 if (!netdev_get_status(iface->netdev, &smap)) {
2183 ovsrec_interface_set_status(iface->cfg, &smap);
2185 ovsrec_interface_set_status(iface->cfg, NULL);
2188 smap_destroy(&smap);
2190 error = netdev_get_flags(iface->netdev, &flags);
2192 const char *state = flags & NETDEV_UP ? "up" : "down";
2194 ovsrec_interface_set_admin_state(iface->cfg, state);
2196 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2199 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2200 ovsrec_interface_set_link_state(iface->cfg, link_state);
2202 link_resets = netdev_get_carrier_resets(iface->netdev);
2203 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2205 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2206 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2208 ovsrec_interface_set_duplex(iface->cfg,
2209 netdev_features_is_full_duplex(current)
2211 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2213 ovsrec_interface_set_duplex(iface->cfg, NULL);
2214 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2217 error = netdev_get_mtu(iface->netdev, &mtu);
2220 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2222 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2225 error = netdev_get_etheraddr(iface->netdev, &mac);
2227 char mac_string[ETH_ADDR_STRLEN + 1];
2229 snprintf(mac_string, sizeof mac_string,
2230 ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2231 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2233 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2236 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2237 * if there is no valid ifindex number. */
2238 ifindex64 = netdev_get_ifindex(iface->netdev);
2239 if (ifindex64 < 0) {
2242 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2246 iface_refresh_ofproto_status(struct iface *iface)
2250 if (iface_is_synthetic(iface)) {
2254 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2258 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2260 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2263 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2265 || status_txn_try_again) {
2266 iface_refresh_cfm_stats(iface);
2269 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2271 || status_txn_try_again) {
2275 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2276 iface->ofp_port, &smap);
2277 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2278 smap_destroy(&smap);
2282 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2285 iface_refresh_cfm_stats(struct iface *iface)
2287 const struct ovsrec_interface *cfg = iface->cfg;
2288 struct cfm_status status;
2291 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2292 iface->ofp_port, &status);
2294 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2295 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2296 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2297 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2298 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2299 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2301 const char *reasons[CFM_FAULT_N_REASONS];
2302 int64_t cfm_health = status.health;
2303 int64_t cfm_flap_count = status.flap_count;
2304 bool faulted = status.faults != 0;
2307 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2310 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2311 int reason = 1 << i;
2312 if (status.faults & reason) {
2313 reasons[j++] = cfm_fault_reason_to_str(reason);
2316 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2318 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2320 if (status.remote_opstate >= 0) {
2321 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2322 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2324 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2327 ovsrec_interface_set_cfm_remote_mpids(cfg,
2328 (const int64_t *)status.rmps,
2330 if (cfm_health >= 0) {
2331 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2333 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2341 iface_refresh_stats(struct iface *iface)
2343 #define IFACE_STATS \
2344 IFACE_STAT(rx_packets, "rx_packets") \
2345 IFACE_STAT(tx_packets, "tx_packets") \
2346 IFACE_STAT(rx_bytes, "rx_bytes") \
2347 IFACE_STAT(tx_bytes, "tx_bytes") \
2348 IFACE_STAT(rx_dropped, "rx_dropped") \
2349 IFACE_STAT(tx_dropped, "tx_dropped") \
2350 IFACE_STAT(rx_errors, "rx_errors") \
2351 IFACE_STAT(tx_errors, "tx_errors") \
2352 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2353 IFACE_STAT(rx_over_errors, "rx_over_err") \
2354 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2355 IFACE_STAT(collisions, "collisions")
2357 #define IFACE_STAT(MEMBER, NAME) + 1
2358 enum { N_IFACE_STATS = IFACE_STATS };
2360 int64_t values[N_IFACE_STATS];
2361 const char *keys[N_IFACE_STATS];
2364 struct netdev_stats stats;
2366 if (iface_is_synthetic(iface)) {
2370 /* Intentionally ignore return value, since errors will set 'stats' to
2371 * all-1s, and we will deal with that correctly below. */
2372 netdev_get_stats(iface->netdev, &stats);
2374 /* Copy statistics into keys[] and values[]. */
2376 #define IFACE_STAT(MEMBER, NAME) \
2377 if (stats.MEMBER != UINT64_MAX) { \
2379 values[n] = stats.MEMBER; \
2384 ovs_assert(n <= N_IFACE_STATS);
2386 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2391 br_refresh_datapath_info(struct bridge *br)
2393 const char *version;
2395 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2396 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2399 ovsrec_bridge_set_datapath_version(br->cfg,
2400 version ? version : "<unknown>");
2404 br_refresh_stp_status(struct bridge *br)
2406 struct smap smap = SMAP_INITIALIZER(&smap);
2407 struct ofproto *ofproto = br->ofproto;
2408 struct ofproto_stp_status status;
2410 if (ofproto_get_stp_status(ofproto, &status)) {
2414 if (!status.enabled) {
2415 ovsrec_bridge_set_status(br->cfg, NULL);
2419 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2420 STP_ID_ARGS(status.bridge_id));
2421 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2422 STP_ID_ARGS(status.designated_root));
2423 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2425 ovsrec_bridge_set_status(br->cfg, &smap);
2426 smap_destroy(&smap);
2430 port_refresh_stp_status(struct port *port)
2432 struct ofproto *ofproto = port->bridge->ofproto;
2433 struct iface *iface;
2434 struct ofproto_port_stp_status status;
2437 if (port_is_synthetic(port)) {
2441 /* STP doesn't currently support bonds. */
2442 if (!list_is_singleton(&port->ifaces)) {
2443 ovsrec_port_set_status(port->cfg, NULL);
2447 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2448 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2452 if (!status.enabled) {
2453 ovsrec_port_set_status(port->cfg, NULL);
2457 /* Set Status column. */
2459 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2460 smap_add(&smap, "stp_state", stp_state_name(status.state));
2461 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2462 smap_add(&smap, "stp_role", stp_role_name(status.role));
2463 ovsrec_port_set_status(port->cfg, &smap);
2464 smap_destroy(&smap);
2468 port_refresh_stp_stats(struct port *port)
2470 struct ofproto *ofproto = port->bridge->ofproto;
2471 struct iface *iface;
2472 struct ofproto_port_stp_stats stats;
2473 const char *keys[3];
2474 int64_t int_values[3];
2476 if (port_is_synthetic(port)) {
2480 /* STP doesn't currently support bonds. */
2481 if (!list_is_singleton(&port->ifaces)) {
2485 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2486 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2490 if (!stats.enabled) {
2491 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2495 /* Set Statistics column. */
2496 keys[0] = "stp_tx_count";
2497 int_values[0] = stats.tx_count;
2498 keys[1] = "stp_rx_count";
2499 int_values[1] = stats.rx_count;
2500 keys[2] = "stp_error_count";
2501 int_values[2] = stats.error_count;
2503 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2504 ARRAY_SIZE(int_values));
2508 br_refresh_rstp_status(struct bridge *br)
2510 struct smap smap = SMAP_INITIALIZER(&smap);
2511 struct ofproto *ofproto = br->ofproto;
2512 struct ofproto_rstp_status status;
2514 if (ofproto_get_rstp_status(ofproto, &status)) {
2517 if (!status.enabled) {
2518 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2521 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2522 RSTP_ID_ARGS(status.bridge_id));
2523 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2524 status.root_path_cost);
2525 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2526 RSTP_ID_ARGS(status.root_id));
2527 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2528 RSTP_ID_ARGS(status.designated_id));
2529 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2530 status.designated_port_id);
2531 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2532 status.bridge_port_id);
2533 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2534 smap_destroy(&smap);
2538 port_refresh_rstp_status(struct port *port)
2540 struct ofproto *ofproto = port->bridge->ofproto;
2541 struct iface *iface;
2542 struct ofproto_port_rstp_status status;
2543 const char *keys[4];
2544 int64_t int_values[4];
2547 if (port_is_synthetic(port)) {
2551 /* RSTP doesn't currently support bonds. */
2552 if (!list_is_singleton(&port->ifaces)) {
2553 ovsrec_port_set_rstp_status(port->cfg, NULL);
2557 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2558 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2562 if (!status.enabled) {
2563 ovsrec_port_set_rstp_status(port->cfg, NULL);
2564 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2567 /* Set Status column. */
2570 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2572 smap_add_format(&smap, "rstp_port_role", "%s",
2573 rstp_port_role_name(status.role));
2574 smap_add_format(&smap, "rstp_port_state", "%s",
2575 rstp_state_name(status.state));
2576 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2577 RSTP_ID_ARGS(status.designated_bridge_id));
2578 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2579 status.designated_port_id);
2580 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2581 status.designated_path_cost);
2583 ovsrec_port_set_rstp_status(port->cfg, &smap);
2584 smap_destroy(&smap);
2586 /* Set Statistics column. */
2587 keys[0] = "rstp_tx_count";
2588 int_values[0] = status.tx_count;
2589 keys[1] = "rstp_rx_count";
2590 int_values[1] = status.rx_count;
2591 keys[2] = "rstp_uptime";
2592 int_values[2] = status.uptime;
2593 keys[3] = "rstp_error_count";
2594 int_values[3] = status.error_count;
2595 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2596 ARRAY_SIZE(int_values));
2600 port_refresh_bond_status(struct port *port, bool force_update)
2602 struct eth_addr mac;
2604 /* Return if port is not a bond */
2605 if (list_is_singleton(&port->ifaces)) {
2609 if (bond_get_changed_active_slave(port->name, &mac, force_update)) {
2613 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2614 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2620 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2622 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2626 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2628 bool enable = enable_system_stats(cfg);
2630 system_stats_enable(enable);
2632 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2637 run_system_stats(void)
2639 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2642 stats = system_stats_run();
2644 struct ovsdb_idl_txn *txn;
2645 struct ovsdb_datum datum;
2647 txn = ovsdb_idl_txn_create(idl);
2648 ovsdb_datum_from_smap(&datum, stats);
2649 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2651 ovsdb_idl_txn_commit(txn);
2652 ovsdb_idl_txn_destroy(txn);
2659 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2662 case OFPCR12_ROLE_EQUAL:
2664 case OFPCR12_ROLE_MASTER:
2666 case OFPCR12_ROLE_SLAVE:
2668 case OFPCR12_ROLE_NOCHANGE:
2670 return "*** INVALID ROLE ***";
2675 refresh_controller_status(void)
2679 const struct ovsrec_controller *cfg;
2683 /* Accumulate status for controllers on all bridges. */
2684 HMAP_FOR_EACH (br, node, &all_bridges) {
2685 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2688 /* Update each controller in the database with current status. */
2689 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2690 struct ofproto_controller_info *cinfo =
2691 shash_find_data(&info, cfg->target);
2694 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2695 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2697 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2699 ovsrec_controller_set_is_connected(cfg, false);
2700 ovsrec_controller_set_role(cfg, NULL);
2701 ovsrec_controller_set_status(cfg, NULL);
2705 ofproto_free_ofproto_controller_info(&info);
2708 /* Update interface and mirror statistics if necessary. */
2710 run_stats_update(void)
2712 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2719 /* Statistics update interval should always be greater than or equal to
2721 stats_interval = MAX(smap_get_int(&cfg->other_config,
2722 "stats-update-interval",
2724 if (stats_timer_interval != stats_interval) {
2725 stats_timer_interval = stats_interval;
2726 stats_timer = LLONG_MIN;
2729 if (time_msec() >= stats_timer) {
2730 enum ovsdb_idl_txn_status status;
2732 /* Rate limit the update. Do not start a new update if the
2733 * previous one is not done. */
2737 stats_txn = ovsdb_idl_txn_create(idl);
2738 HMAP_FOR_EACH (br, node, &all_bridges) {
2742 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2743 struct iface *iface;
2745 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2746 iface_refresh_stats(iface);
2748 port_refresh_stp_stats(port);
2750 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2751 mirror_refresh_stats(m);
2754 refresh_controller_status();
2757 status = ovsdb_idl_txn_commit(stats_txn);
2758 if (status != TXN_INCOMPLETE) {
2759 stats_timer = time_msec() + stats_timer_interval;
2760 ovsdb_idl_txn_destroy(stats_txn);
2767 stats_update_wait(void)
2769 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2770 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2772 ovsdb_idl_txn_wait(stats_txn);
2774 poll_timer_wait_until(stats_timer);
2778 /* Update bridge/port/interface status if necessary. */
2780 run_status_update(void)
2785 /* Rate limit the update. Do not start a new update if the
2786 * previous one is not done. */
2787 seq = seq_read(connectivity_seq_get());
2788 if (seq != connectivity_seqno || status_txn_try_again) {
2791 connectivity_seqno = seq;
2792 status_txn = ovsdb_idl_txn_create(idl);
2793 HMAP_FOR_EACH (br, node, &all_bridges) {
2796 br_refresh_stp_status(br);
2797 br_refresh_rstp_status(br);
2798 br_refresh_datapath_info(br);
2799 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2800 struct iface *iface;
2802 port_refresh_stp_status(port);
2803 port_refresh_rstp_status(port);
2804 port_refresh_bond_status(port, status_txn_try_again);
2805 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2806 iface_refresh_netdev_status(iface);
2807 iface_refresh_ofproto_status(iface);
2814 /* Commit the transaction and get the status. If the transaction finishes,
2815 * then destroy the transaction. Otherwise, keep it so that we can check
2816 * progress the next time that this function is called. */
2818 enum ovsdb_idl_txn_status status;
2820 status = ovsdb_idl_txn_commit(status_txn);
2821 if (status != TXN_INCOMPLETE) {
2822 ovsdb_idl_txn_destroy(status_txn);
2825 /* Sets the 'status_txn_try_again' if the transaction fails. */
2826 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2827 status_txn_try_again = false;
2829 status_txn_try_again = true;
2834 /* Refresh AA port status if necessary. */
2835 if (time_msec() >= aa_refresh_timer) {
2838 HMAP_FOR_EACH (br, node, &all_bridges) {
2839 if (bridge_aa_need_refresh(br)) {
2840 struct ovsdb_idl_txn *txn;
2842 txn = ovsdb_idl_txn_create(idl);
2843 bridge_aa_refresh_queued(br);
2844 ovsdb_idl_txn_commit(txn);
2845 ovsdb_idl_txn_destroy(txn);
2849 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2854 status_update_wait(void)
2856 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2857 * transaction to complete. If the status update to database needs to be
2858 * run again (transaction fails), registers a timeout in
2859 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2860 * sequence number. */
2862 ovsdb_idl_txn_wait(status_txn);
2863 } else if (status_txn_try_again) {
2864 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2866 seq_wait(connectivity_seq_get(), connectivity_seqno);
2877 /* Let each datapath type do the work that it needs to do. */
2879 ofproto_enumerate_types(&types);
2880 SSET_FOR_EACH (type, &types) {
2881 ofproto_type_run(type);
2883 sset_destroy(&types);
2885 /* Let each bridge do the work that it needs to do. */
2886 HMAP_FOR_EACH (br, node, &all_bridges) {
2887 ofproto_run(br->ofproto);
2894 static struct ovsrec_open_vswitch null_cfg;
2895 const struct ovsrec_open_vswitch *cfg;
2897 bool vlan_splinters_changed;
2899 ovsrec_open_vswitch_init(&null_cfg);
2905 if (ovsdb_idl_is_lock_contended(idl)) {
2906 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2907 struct bridge *br, *next_br;
2909 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2910 "disabling this process (pid %ld) until it goes away",
2911 (long int) getpid());
2913 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2916 /* Since we will not be running system_stats_run() in this process
2917 * with the current situation of multiple ovs-vswitchd daemons,
2918 * disable system stats collection. */
2919 system_stats_enable(false);
2921 } else if (!ovsdb_idl_has_lock(idl)
2922 || !ovsdb_idl_has_ever_connected(idl)) {
2923 /* Returns if not holding the lock or not done retrieving db
2927 cfg = ovsrec_open_vswitch_first(idl);
2929 /* Initialize the ofproto library. This only needs to run once, but
2930 * it must be done after the configuration is set. If the
2931 * initialization has already occurred, bridge_init_ofproto()
2932 * returns immediately. */
2933 bridge_init_ofproto(cfg);
2935 /* Once the value of flow-restore-wait is false, we no longer should
2936 * check its value from the database. */
2937 if (cfg && ofproto_get_flow_restore_wait()) {
2938 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2939 "flow-restore-wait", false));
2944 /* Re-configure SSL. We do this on every trip through the main loop,
2945 * instead of just when the database changes, because the contents of the
2946 * key and certificate files can change without the database changing.
2948 * We do this before bridge_reconfigure() because that function might
2949 * initiate SSL connections and thus requires SSL to be configured. */
2950 if (cfg && cfg->ssl) {
2951 const struct ovsrec_ssl *ssl = cfg->ssl;
2953 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2954 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2957 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2958 * usage has changed. */
2959 vlan_splinters_changed = false;
2960 if (vlan_splinters_enabled_anywhere) {
2963 HMAP_FOR_EACH (br, node, &all_bridges) {
2964 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2965 vlan_splinters_changed = true;
2971 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed
2972 || ifaces_changed) {
2973 struct ovsdb_idl_txn *txn;
2975 ifaces_changed = false;
2977 idl_seqno = ovsdb_idl_get_seqno(idl);
2978 txn = ovsdb_idl_txn_create(idl);
2979 bridge_reconfigure(cfg ? cfg : &null_cfg);
2982 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2983 discover_types(cfg);
2986 /* If we are completing our initial configuration for this run
2987 * of ovs-vswitchd, then keep the transaction around to monitor
2988 * it for completion. */
2989 if (initial_config_done) {
2990 /* Always sets the 'status_txn_try_again' to check again,
2991 * in case that this transaction fails. */
2992 status_txn_try_again = true;
2993 ovsdb_idl_txn_commit(txn);
2994 ovsdb_idl_txn_destroy(txn);
2996 initial_config_done = true;
2997 daemonize_txn = txn;
3001 if (daemonize_txn) {
3002 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
3003 if (status != TXN_INCOMPLETE) {
3004 ovsdb_idl_txn_destroy(daemonize_txn);
3005 daemonize_txn = NULL;
3007 /* ovs-vswitchd has completed initialization, so allow the
3008 * process that forked us to exit successfully. */
3009 daemonize_complete();
3011 vlog_enable_async();
3013 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3018 run_status_update();
3028 ovsdb_idl_wait(idl);
3029 if (daemonize_txn) {
3030 ovsdb_idl_txn_wait(daemonize_txn);
3034 if (ifaces_changed) {
3035 poll_immediate_wake();
3039 ofproto_enumerate_types(&types);
3040 SSET_FOR_EACH (type, &types) {
3041 ofproto_type_wait(type);
3043 sset_destroy(&types);
3045 if (!hmap_is_empty(&all_bridges)) {
3048 HMAP_FOR_EACH (br, node, &all_bridges) {
3049 ofproto_wait(br->ofproto);
3051 stats_update_wait();
3052 status_update_wait();
3055 system_stats_wait();
3058 /* Adds some memory usage statistics for bridges into 'usage', for use with
3059 * memory_report(). */
3061 bridge_get_memory_usage(struct simap *usage)
3068 ofproto_enumerate_types(&types);
3069 SSET_FOR_EACH (type, &types) {
3070 ofproto_type_get_memory_usage(type, usage);
3072 sset_destroy(&types);
3074 HMAP_FOR_EACH (br, node, &all_bridges) {
3075 ofproto_get_memory_usage(br->ofproto, usage);
3079 /* QoS unixctl user interface functions. */
3081 struct qos_unixctl_show_cbdata {
3083 struct iface *iface;
3087 qos_unixctl_show_queue(unsigned int queue_id,
3088 const struct smap *details,
3089 struct iface *iface,
3092 struct netdev_queue_stats stats;
3093 struct smap_node *node;
3096 ds_put_cstr(ds, "\n");
3098 ds_put_format(ds, "Queue %u:\n", queue_id);
3100 ds_put_cstr(ds, "Default:\n");
3103 SMAP_FOR_EACH (node, details) {
3104 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3107 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3109 if (stats.tx_packets != UINT64_MAX) {
3110 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3113 if (stats.tx_bytes != UINT64_MAX) {
3114 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3117 if (stats.tx_errors != UINT64_MAX) {
3118 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3121 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3122 queue_id, ovs_strerror(error));
3127 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3128 const char *argv[], void *aux OVS_UNUSED)
3130 struct ds ds = DS_EMPTY_INITIALIZER;
3131 struct smap smap = SMAP_INITIALIZER(&smap);
3132 struct iface *iface;
3134 struct smap_node *node;
3136 iface = iface_find(argv[1]);
3138 unixctl_command_reply_error(conn, "no such interface");
3142 netdev_get_qos(iface->netdev, &type, &smap);
3144 if (*type != '\0') {
3145 struct netdev_queue_dump dump;
3146 struct smap details;
3147 unsigned int queue_id;
3149 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3151 SMAP_FOR_EACH (node, &smap) {
3152 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3155 smap_init(&details);
3156 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3157 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3159 smap_destroy(&details);
3161 unixctl_command_reply(conn, ds_cstr(&ds));
3163 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3164 unixctl_command_reply_error(conn, ds_cstr(&ds));
3167 smap_destroy(&smap);
3171 /* Bridge reconfiguration functions. */
3173 bridge_create(const struct ovsrec_bridge *br_cfg)
3177 ovs_assert(!bridge_lookup(br_cfg->name));
3178 br = xzalloc(sizeof *br);
3180 br->name = xstrdup(br_cfg->name);
3181 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3184 /* Derive the default Ethernet address from the bridge's UUID. This should
3185 * be unique and it will be stable between ovs-vswitchd runs. */
3186 memcpy(&br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3187 eth_addr_mark_random(&br->default_ea);
3189 hmap_init(&br->ports);
3190 hmap_init(&br->ifaces);
3191 hmap_init(&br->iface_by_name);
3192 hmap_init(&br->mirrors);
3194 hmap_init(&br->mappings);
3195 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3199 bridge_destroy(struct bridge *br)
3202 struct mirror *mirror, *next_mirror;
3203 struct port *port, *next_port;
3205 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3208 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3209 mirror_destroy(mirror);
3212 hmap_remove(&all_bridges, &br->node);
3213 ofproto_destroy(br->ofproto);
3214 hmap_destroy(&br->ifaces);
3215 hmap_destroy(&br->ports);
3216 hmap_destroy(&br->iface_by_name);
3217 hmap_destroy(&br->mirrors);
3218 hmap_destroy(&br->mappings);
3225 static struct bridge *
3226 bridge_lookup(const char *name)
3230 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3231 if (!strcmp(br->name, name)) {
3238 /* Handle requests for a listing of all flows known by the OpenFlow
3239 * stack, including those normally hidden. */
3241 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3242 const char *argv[], void *aux OVS_UNUSED)
3247 br = bridge_lookup(argv[1]);
3249 unixctl_command_reply_error(conn, "Unknown bridge");
3254 ofproto_get_all_flows(br->ofproto, &results);
3256 unixctl_command_reply(conn, ds_cstr(&results));
3257 ds_destroy(&results);
3260 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3261 * connections and reconnect. If BRIDGE is not specified, then all bridges
3262 * drop their controller connections and reconnect. */
3264 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3265 const char *argv[], void *aux OVS_UNUSED)
3269 br = bridge_lookup(argv[1]);
3271 unixctl_command_reply_error(conn, "Unknown bridge");
3274 ofproto_reconnect_controllers(br->ofproto);
3276 HMAP_FOR_EACH (br, node, &all_bridges) {
3277 ofproto_reconnect_controllers(br->ofproto);
3280 unixctl_command_reply(conn, NULL);
3284 bridge_get_controllers(const struct bridge *br,
3285 struct ovsrec_controller ***controllersp)
3287 struct ovsrec_controller **controllers;
3288 size_t n_controllers;
3290 controllers = br->cfg->controller;
3291 n_controllers = br->cfg->n_controller;
3293 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3299 *controllersp = controllers;
3301 return n_controllers;
3305 bridge_collect_wanted_ports(struct bridge *br,
3306 const unsigned long int *splinter_vlans,
3307 struct shash *wanted_ports)
3311 shash_init(wanted_ports);
3313 for (i = 0; i < br->cfg->n_ports; i++) {
3314 const char *name = br->cfg->ports[i]->name;
3315 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3316 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3320 if (bridge_get_controllers(br, NULL)
3321 && !shash_find(wanted_ports, br->name)) {
3322 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3323 br->name, br->name);
3325 ovsrec_interface_init(&br->synth_local_iface);
3326 ovsrec_port_init(&br->synth_local_port);
3328 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3329 br->synth_local_port.n_interfaces = 1;
3330 br->synth_local_port.name = br->name;
3332 br->synth_local_iface.name = br->name;
3333 br->synth_local_iface.type = "internal";
3335 br->synth_local_ifacep = &br->synth_local_iface;
3337 shash_add(wanted_ports, br->name, &br->synth_local_port);
3340 if (splinter_vlans) {
3341 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3345 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3346 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3347 * 'br' needs to complete its configuration. */
3349 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3351 struct shash_node *port_node;
3352 struct port *port, *next;
3354 /* Get rid of deleted ports.
3355 * Get rid of deleted interfaces on ports that still exist. */
3356 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3357 port->cfg = shash_find_data(wanted_ports, port->name);
3361 port_del_ifaces(port);
3365 /* Update iface->cfg and iface->type in interfaces that still exist. */
3366 SHASH_FOR_EACH (port_node, wanted_ports) {
3367 const struct ovsrec_port *port = port_node->data;
3370 for (i = 0; i < port->n_interfaces; i++) {
3371 const struct ovsrec_interface *cfg = port->interfaces[i];
3372 struct iface *iface = iface_lookup(br, cfg->name);
3373 const char *type = iface_get_type(cfg, br->cfg);
3378 } else if (!strcmp(type, "null")) {
3379 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3380 " may be removed in February 2013. Please email"
3381 " dev@openvswitch.org with concerns.",
3384 /* We will add new interfaces later. */
3390 /* Initializes 'oc' appropriately as a management service controller for
3393 * The caller must free oc->target when it is no longer needed. */
3395 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3396 struct ofproto_controller *oc)
3398 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3399 oc->max_backoff = 0;
3400 oc->probe_interval = 60;
3401 oc->band = OFPROTO_OUT_OF_BAND;
3403 oc->burst_limit = 0;
3404 oc->enable_async_msgs = true;
3408 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3410 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3411 struct ofproto_controller *oc)
3415 oc->target = c->target;
3416 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3417 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3418 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3419 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3420 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3421 oc->burst_limit = (c->controller_burst_limit
3422 ? *c->controller_burst_limit : 0);
3423 oc->enable_async_msgs = (!c->enable_async_messages
3424 || *c->enable_async_messages);
3425 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3426 if (dscp < 0 || dscp > 63) {
3427 dscp = DSCP_DEFAULT;
3432 /* Configures the IP stack for 'br''s local interface properly according to the
3433 * configuration in 'c'. */
3435 bridge_configure_local_iface_netdev(struct bridge *br,
3436 struct ovsrec_controller *c)
3438 struct netdev *netdev;
3439 struct in_addr mask, gateway;
3441 struct iface *local_iface;
3444 /* If there's no local interface or no IP address, give up. */
3445 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3446 if (!local_iface || !c->local_ip || !ip_parse(c->local_ip, &ip.s_addr)) {
3450 /* Bring up the local interface. */
3451 netdev = local_iface->netdev;
3452 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3454 /* Configure the IP address and netmask. */
3455 if (!c->local_netmask
3456 || !ip_parse(c->local_netmask, &mask.s_addr)
3458 mask.s_addr = guess_netmask(ip.s_addr);
3460 if (!netdev_set_in4(netdev, ip, mask)) {
3461 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3462 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3465 /* Configure the default gateway. */
3466 if (c->local_gateway
3467 && ip_parse(c->local_gateway, &gateway.s_addr)
3468 && gateway.s_addr) {
3469 if (!netdev_add_router(netdev, gateway)) {
3470 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3471 br->name, IP_ARGS(gateway.s_addr));
3476 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3477 * in either string are treated as equal to any number of slashes in the other,
3478 * e.g. "x///y" is equal to "x/y".
3480 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3481 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3482 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3483 * 'b' against a prefix of 'a'.
3486 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3488 const char *b_start = b;
3490 if (b - b_start >= b_stoplen) {
3492 } else if (*a != *b) {
3494 } else if (*a == '/') {
3495 a += strspn(a, "/");
3496 b += strspn(b, "/");
3497 } else if (*a == '\0') {
3507 bridge_configure_remotes(struct bridge *br,
3508 const struct sockaddr_in *managers, size_t n_managers)
3510 bool disable_in_band;
3512 struct ovsrec_controller **controllers;
3513 size_t n_controllers;
3515 enum ofproto_fail_mode fail_mode;
3517 struct ofproto_controller *ocs;
3521 /* Check if we should disable in-band control on this bridge. */
3522 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3525 /* Set OpenFlow queue ID for in-band control. */
3526 ofproto_set_in_band_queue(br->ofproto,
3527 smap_get_int(&br->cfg->other_config,
3528 "in-band-queue", -1));
3530 if (disable_in_band) {
3531 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3533 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3536 n_controllers = bridge_get_controllers(br, &controllers);
3538 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3541 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3542 for (i = 0; i < n_controllers; i++) {
3543 struct ovsrec_controller *c = controllers[i];
3545 if (!strncmp(c->target, "punix:", 6)
3546 || !strncmp(c->target, "unix:", 5)) {
3547 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3550 if (!strncmp(c->target, "unix:", 5)) {
3551 /* Connect to a listening socket */
3552 whitelist = xasprintf("unix:%s/", ovs_rundir());
3553 if (strchr(c->target, '/') &&
3554 !equal_pathnames(c->target, whitelist,
3555 strlen(whitelist))) {
3556 /* Absolute path specified, but not in ovs_rundir */
3557 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3558 "controller \"%s\" due to possibility for "
3559 "remote exploit. Instead, specify socket "
3560 "in whitelisted \"%s\" or connect to "
3561 "\"unix:%s/%s.mgmt\" (which is always "
3562 "available without special configuration).",
3563 br->name, c->target, whitelist,
3564 ovs_rundir(), br->name);
3569 whitelist = xasprintf("punix:%s/%s.",
3570 ovs_rundir(), br->name);
3571 if (!equal_pathnames(c->target, whitelist, strlen(whitelist))
3572 || strchr(c->target + strlen(whitelist), '/')) {
3573 /* Prevent remote ovsdb-server users from accessing
3574 * arbitrary Unix domain sockets and overwriting arbitrary
3576 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3577 "controller \"%s\" due to possibility of "
3578 "overwriting local files. Instead, specify "
3579 "path in whitelisted format \"%s*\" or "
3580 "connect to \"unix:%s/%s.mgmt\" (which is "
3581 "always available without special "
3583 br->name, c->target, whitelist,
3584 ovs_rundir(), br->name);
3593 bridge_configure_local_iface_netdev(br, c);
3594 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3595 if (disable_in_band) {
3596 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3601 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3602 bridge_get_allowed_versions(br));
3603 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3606 /* Set the fail-mode. */
3607 fail_mode = !br->cfg->fail_mode
3608 || !strcmp(br->cfg->fail_mode, "standalone")
3609 ? OFPROTO_FAIL_STANDALONE
3610 : OFPROTO_FAIL_SECURE;
3611 ofproto_set_fail_mode(br->ofproto, fail_mode);
3613 /* Configure OpenFlow controller connection snooping. */
3614 if (!ofproto_has_snoops(br->ofproto)) {
3618 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3619 ovs_rundir(), br->name));
3620 ofproto_set_snoops(br->ofproto, &snoops);
3621 sset_destroy(&snoops);
3626 bridge_configure_tables(struct bridge *br)
3628 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3632 n_tables = ofproto_get_n_tables(br->ofproto);
3634 for (i = 0; i < n_tables; i++) {
3635 struct ofproto_table_settings s;
3636 bool use_default_prefixes = true;
3639 s.max_flows = UINT_MAX;
3641 s.enable_eviction = false;
3643 s.n_prefix_fields = 0;
3644 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3646 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3647 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3650 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3651 s.max_flows = *cfg->flow_limit;
3654 s.enable_eviction = (cfg->overflow_policy
3655 && !strcmp(cfg->overflow_policy, "evict"));
3656 if (cfg->n_groups) {
3657 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3658 for (k = 0; k < cfg->n_groups; k++) {
3659 const char *string = cfg->groups[k];
3662 msg = mf_parse_subfield__(&s.groups[k], &string);
3664 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3665 "'groups' (%s)", br->name, i, msg);
3667 } else if (*string) {
3668 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3669 "element '%s' contains trailing garbage",
3670 br->name, i, cfg->groups[k]);
3677 /* Prefix lookup fields. */
3678 s.n_prefix_fields = 0;
3679 for (k = 0; k < cfg->n_prefixes; k++) {
3680 const char *name = cfg->prefixes[k];
3681 const struct mf_field *mf;
3683 if (strcmp(name, "none") == 0) {
3684 use_default_prefixes = false;
3685 s.n_prefix_fields = 0;
3688 mf = mf_from_name(name);
3690 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3694 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3695 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3696 "%s", br->name, name);
3699 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3700 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3701 "field not used: %s", br->name, name);
3704 use_default_prefixes = false;
3705 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3708 if (use_default_prefixes) {
3709 /* Use default values. */
3710 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3711 memcpy(s.prefix_fields, default_prefix_fields,
3712 sizeof default_prefix_fields);
3715 struct ds ds = DS_EMPTY_INITIALIZER;
3716 for (k = 0; k < s.n_prefix_fields; k++) {
3718 ds_put_char(&ds, ',');
3720 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3722 if (s.n_prefix_fields == 0) {
3723 ds_put_cstr(&ds, "none");
3725 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3726 br->name, i, ds_cstr(&ds));
3730 ofproto_configure_table(br->ofproto, i, &s);
3734 for (; j < br->cfg->n_flow_tables; j++) {
3735 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3736 "%"PRId64" not supported by this datapath", br->name,
3737 br->cfg->key_flow_tables[j]);
3742 bridge_configure_dp_desc(struct bridge *br)
3744 ofproto_set_dp_desc(br->ofproto,
3745 smap_get(&br->cfg->other_config, "dp-desc"));
3748 static struct aa_mapping *
3749 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3751 struct aa_mapping *m;
3753 HMAP_FOR_EACH_IN_BUCKET (m,
3755 hash_bytes(&isid, sizeof isid, 0),
3757 if (isid == m->isid) {
3764 static struct aa_mapping *
3765 bridge_aa_mapping_create(struct bridge *br,
3769 struct aa_mapping *m;
3771 m = xzalloc(sizeof *m);
3775 m->br_name = xstrdup(br->name);
3776 hmap_insert(&br->mappings,
3778 hash_bytes(&isid, sizeof isid, 0));
3784 bridge_aa_mapping_destroy(struct aa_mapping *m)
3787 struct bridge *br = m->bridge;
3790 ofproto_aa_mapping_unregister(br->ofproto, m);
3793 hmap_remove(&br->mappings, &m->hmap_node);
3802 bridge_aa_mapping_configure(struct aa_mapping *m)
3804 struct aa_mapping_settings s;
3810 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3816 bridge_configure_aa(struct bridge *br)
3818 const struct ovsdb_datum *mc;
3819 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3820 struct aa_settings aa_s;
3821 struct aa_mapping *m, *next;
3825 ofproto_set_aa(br->ofproto, NULL, NULL);
3829 memset(&aa_s, 0, sizeof aa_s);
3830 aa_s.system_description = auto_attach->system_description;
3831 aa_s.system_name = auto_attach->system_name;
3832 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3834 mc = ovsrec_autoattach_get_mappings(auto_attach,
3836 OVSDB_TYPE_INTEGER);
3837 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3838 union ovsdb_atom atom;
3840 atom.integer = m->isid;
3841 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3842 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3844 bridge_aa_mapping_destroy(m);
3848 /* Add new mappings and reconfigure existing ones. */
3849 for (i = 0; i < auto_attach->n_mappings; ++i) {
3850 struct aa_mapping *m =
3851 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3854 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3855 auto_attach->key_mappings[i],
3856 auto_attach->value_mappings[i]);
3857 m = bridge_aa_mapping_create(br,
3858 auto_attach->key_mappings[i],
3859 auto_attach->value_mappings[i]);
3861 if (!bridge_aa_mapping_configure(m)) {
3862 bridge_aa_mapping_destroy(m);
3869 bridge_aa_need_refresh(struct bridge *br)
3871 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3875 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3877 int64_t *trunks = NULL;
3879 bool found = false, reconfigure = false;
3881 for (i = 0; i < port->cfg->n_trunks; i++) {
3882 if (port->cfg->trunks[i] == m->vlan) {
3889 case BRIDGE_AA_VLAN_OPER_ADD:
3891 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3893 for (i = 0; i < port->cfg->n_trunks; i++) {
3894 trunks[i] = port->cfg->trunks[i];
3896 trunks[i++] = m->vlan;
3902 case BRIDGE_AA_VLAN_OPER_REMOVE:
3906 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3908 for (i = 0; i < port->cfg->n_trunks; i++) {
3909 if (port->cfg->trunks[i] != m->vlan) {
3910 trunks[j++] = port->cfg->trunks[i];
3919 case BRIDGE_AA_VLAN_OPER_UNDEF:
3921 VLOG_WARN("unrecognized operation %u", m->oper);
3926 /* VLAN switching under trunk mode cause the trunk port to switch all
3927 * VLANs, see ovs-vswitchd.conf.db
3930 static char *vlan_mode_access = "access";
3931 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3935 static char *vlan_mode_trunk = "trunk";
3936 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3939 ovsrec_port_set_trunks(port->cfg, trunks, i);
3941 /* Force reconfigure of the port. */
3942 port_configure(port);
3947 bridge_aa_refresh_queued(struct bridge *br)
3949 struct ovs_list *list = xmalloc(sizeof *list);
3950 struct bridge_aa_vlan *node, *next;
3953 ofproto_aa_vlan_get_queued(br->ofproto, list);
3955 LIST_FOR_EACH_SAFE (node, next, list_node, list) {
3958 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3961 port = port_lookup(br, node->port_name);
3963 bridge_aa_update_trunks(port, node);
3966 list_remove(&node->list_node);
3967 free(node->port_name);
3975 /* Port functions. */
3977 static struct port *
3978 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3982 port = xzalloc(sizeof *port);
3984 port->name = xstrdup(cfg->name);
3986 list_init(&port->ifaces);
3988 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3992 /* Deletes interfaces from 'port' that are no longer configured for it. */
3994 port_del_ifaces(struct port *port)
3996 struct iface *iface, *next;
3997 struct sset new_ifaces;
4000 /* Collect list of new interfaces. */
4001 sset_init(&new_ifaces);
4002 for (i = 0; i < port->cfg->n_interfaces; i++) {
4003 const char *name = port->cfg->interfaces[i]->name;
4004 const char *type = port->cfg->interfaces[i]->type;
4005 if (strcmp(type, "null")) {
4006 sset_add(&new_ifaces, name);
4010 /* Get rid of deleted interfaces. */
4011 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4012 if (!sset_contains(&new_ifaces, iface->name)) {
4013 iface_destroy(iface);
4017 sset_destroy(&new_ifaces);
4021 port_destroy(struct port *port)
4024 struct bridge *br = port->bridge;
4025 struct iface *iface, *next;
4028 ofproto_bundle_unregister(br->ofproto, port);
4031 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4032 iface_destroy__(iface);
4035 hmap_remove(&br->ports, &port->hmap_node);
4041 static struct port *
4042 port_lookup(const struct bridge *br, const char *name)
4046 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4048 if (!strcmp(port->name, name)) {
4056 enable_lacp(struct port *port, bool *activep)
4058 if (!port->cfg->lacp) {
4059 /* XXX when LACP implementation has been sufficiently tested, enable by
4060 * default and make active on bonded ports. */
4062 } else if (!strcmp(port->cfg->lacp, "off")) {
4064 } else if (!strcmp(port->cfg->lacp, "active")) {
4067 } else if (!strcmp(port->cfg->lacp, "passive")) {
4071 VLOG_WARN("port %s: unknown LACP mode %s",
4072 port->name, port->cfg->lacp);
4077 static struct lacp_settings *
4078 port_configure_lacp(struct port *port, struct lacp_settings *s)
4080 const char *lacp_time, *system_id;
4083 if (!enable_lacp(port, &s->active)) {
4087 s->name = port->name;
4089 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4091 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4092 ETH_ADDR_SCAN_ARGS(s->id))) {
4093 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4094 " address.", port->name, system_id);
4098 s->id = port->bridge->ea;
4101 if (eth_addr_is_zero(s->id)) {
4102 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4106 /* Prefer bondable links if unspecified. */
4107 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4109 s->priority = (priority > 0 && priority <= UINT16_MAX
4111 : UINT16_MAX - !list_is_short(&port->ifaces));
4113 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4114 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4116 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4117 "lacp-fallback-ab", false);
4123 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4125 int priority, portid, key;
4127 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4128 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4130 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4132 if (portid <= 0 || portid > UINT16_MAX) {
4133 portid = ofp_to_u16(iface->ofp_port);
4136 if (priority <= 0 || priority > UINT16_MAX) {
4137 priority = UINT16_MAX;
4140 if (key < 0 || key > UINT16_MAX) {
4144 s->name = iface->name;
4146 s->priority = priority;
4151 port_configure_bond(struct port *port, struct bond_settings *s)
4153 const char *detect_s;
4154 struct iface *iface;
4156 int miimon_interval;
4158 s->name = port->name;
4160 if (port->cfg->bond_mode) {
4161 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4162 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4163 port->name, port->cfg->bond_mode,
4164 bond_mode_to_string(s->balance));
4167 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4169 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4170 * active-backup. At some point we should remove this warning. */
4171 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4172 " in previous versions, the default bond_mode was"
4173 " balance-slb", port->name,
4174 bond_mode_to_string(s->balance));
4176 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4177 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4178 "please use another bond type or disable flood_vlans",
4182 miimon_interval = smap_get_int(&port->cfg->other_config,
4183 "bond-miimon-interval", 0);
4184 if (miimon_interval <= 0) {
4185 miimon_interval = 200;
4188 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4189 if (!detect_s || !strcmp(detect_s, "carrier")) {
4190 miimon_interval = 0;
4191 } else if (strcmp(detect_s, "miimon")) {
4192 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4193 "defaulting to carrier", port->name, detect_s);
4194 miimon_interval = 0;
4197 s->up_delay = MAX(0, port->cfg->bond_updelay);
4198 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4199 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4200 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4201 "bond-rebalance-interval", 10000);
4202 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4203 s->rebalance_interval = 1000;
4206 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4207 "lacp-fallback-ab", false);
4209 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4210 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4213 mac_s = port->cfg->bond_active_slave;
4214 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4215 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4216 /* OVSDB did not store the last active interface */
4217 s->active_slave_mac = eth_addr_zero;
4221 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4222 * instead of obtaining it from the database. */
4224 port_is_synthetic(const struct port *port)
4226 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4229 /* Interface functions. */
4232 iface_is_internal(const struct ovsrec_interface *iface,
4233 const struct ovsrec_bridge *br)
4235 /* The local port and "internal" ports are always "internal". */
4236 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4239 /* Returns the correct network device type for interface 'iface' in bridge
4242 iface_get_type(const struct ovsrec_interface *iface,
4243 const struct ovsrec_bridge *br)
4247 /* The local port always has type "internal". Other ports take
4248 * their type from the database and default to "system" if none is
4250 if (iface_is_internal(iface, br)) {
4253 type = iface->type[0] ? iface->type : "system";
4256 return ofproto_port_open_type(br->datapath_type, type);
4260 iface_destroy__(struct iface *iface)
4263 struct port *port = iface->port;
4264 struct bridge *br = port->bridge;
4266 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4267 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4270 if (iface->ofp_port != OFPP_NONE) {
4271 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4274 list_remove(&iface->port_elem);
4275 hmap_remove(&br->iface_by_name, &iface->name_node);
4277 /* The user is changing configuration here, so netdev_remove needs to be
4278 * used as opposed to netdev_close */
4279 netdev_remove(iface->netdev);
4287 iface_destroy(struct iface *iface)
4290 struct port *port = iface->port;
4292 iface_destroy__(iface);
4293 if (list_is_empty(&port->ifaces)) {
4299 static struct iface *
4300 iface_lookup(const struct bridge *br, const char *name)
4302 struct iface *iface;
4304 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4305 &br->iface_by_name) {
4306 if (!strcmp(iface->name, name)) {
4314 static struct iface *
4315 iface_find(const char *name)
4317 const struct bridge *br;
4319 HMAP_FOR_EACH (br, node, &all_bridges) {
4320 struct iface *iface = iface_lookup(br, name);
4329 static struct iface *
4330 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4332 struct iface *iface;
4334 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4336 if (iface->ofp_port == ofp_port) {
4343 /* Set Ethernet address of 'iface', if one is specified in the configuration
4346 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4348 struct eth_addr ea, *mac = NULL;
4349 struct iface *hw_addr_iface;
4351 if (strcmp(iface->type, "internal")) {
4355 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, &ea)) {
4357 } else if (port->cfg->fake_bridge) {
4358 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4359 find_local_hw_addr(br, &ea, port, &hw_addr_iface);
4364 if (iface->ofp_port == OFPP_LOCAL) {
4365 VLOG_ERR("interface %s: ignoring mac in Interface record "
4366 "(use Bridge record to set local port's mac)",
4368 } else if (eth_addr_is_multicast(*mac)) {
4369 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4372 int error = netdev_set_etheraddr(iface->netdev, *mac);
4374 VLOG_ERR("interface %s: setting MAC failed (%s)",
4375 iface->name, ovs_strerror(error));
4381 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4383 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4385 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4386 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4387 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4391 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4392 * sets the "ofport" field to -1.
4394 * This is appropriate when 'if_cfg''s interface cannot be created or is
4395 * otherwise invalid. */
4397 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4399 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4400 iface_set_ofport(if_cfg, OFPP_NONE);
4401 ovsrec_interface_set_error(if_cfg, errp);
4402 ovsrec_interface_set_status(if_cfg, NULL);
4403 ovsrec_interface_set_admin_state(if_cfg, NULL);
4404 ovsrec_interface_set_duplex(if_cfg, NULL);
4405 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4406 ovsrec_interface_set_link_state(if_cfg, NULL);
4407 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4408 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4409 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4410 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4411 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4412 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4413 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4414 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4419 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4421 union ovsdb_atom atom;
4423 atom.integer = target;
4424 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4428 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4430 struct ofpbuf queues_buf;
4432 ofpbuf_init(&queues_buf, 0);
4434 if (!qos || qos->type[0] == '\0') {
4435 netdev_set_qos(iface->netdev, NULL, NULL);
4437 const struct ovsdb_datum *queues;
4438 struct netdev_queue_dump dump;
4439 unsigned int queue_id;
4440 struct smap details;
4444 /* Configure top-level Qos for 'iface'. */
4445 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4447 /* Deconfigure queues that were deleted. */
4448 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4450 smap_init(&details);
4451 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4452 if (!queue_ids_include(queues, queue_id)) {
4453 netdev_delete_queue(iface->netdev, queue_id);
4456 smap_destroy(&details);
4458 /* Configure queues for 'iface'. */
4460 for (i = 0; i < qos->n_queues; i++) {
4461 const struct ovsrec_queue *queue = qos->value_queues[i];
4462 unsigned int queue_id = qos->key_queues[i];
4464 if (queue_id == 0) {
4468 if (queue->n_dscp == 1) {
4469 struct ofproto_port_queue *port_queue;
4471 port_queue = ofpbuf_put_uninit(&queues_buf,
4472 sizeof *port_queue);
4473 port_queue->queue = queue_id;
4474 port_queue->dscp = queue->dscp[0];
4477 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4480 struct smap details;
4482 smap_init(&details);
4483 netdev_set_queue(iface->netdev, 0, &details);
4484 smap_destroy(&details);
4488 if (iface->ofp_port != OFPP_NONE) {
4489 const struct ofproto_port_queue *port_queues = queues_buf.data;
4490 size_t n_queues = queues_buf.size / sizeof *port_queues;
4492 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4493 port_queues, n_queues);
4496 netdev_set_policing(iface->netdev,
4497 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4498 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4500 ofpbuf_uninit(&queues_buf);
4504 iface_configure_cfm(struct iface *iface)
4506 const struct ovsrec_interface *cfg = iface->cfg;
4507 const char *opstate_str;
4508 const char *cfm_ccm_vlan;
4509 struct cfm_settings s;
4510 struct smap netdev_args;
4512 if (!cfg->n_cfm_mpid) {
4513 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4517 s.check_tnl_key = false;
4518 smap_init(&netdev_args);
4519 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4520 const char *key = smap_get(&netdev_args, "key");
4521 const char *in_key = smap_get(&netdev_args, "in_key");
4523 s.check_tnl_key = (key && !strcmp(key, "flow"))
4524 || (in_key && !strcmp(in_key, "flow"));
4526 smap_destroy(&netdev_args);
4528 s.mpid = *cfg->cfm_mpid;
4529 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4530 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4531 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4533 if (s.interval <= 0) {
4537 if (!cfm_ccm_vlan) {
4539 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4540 s.ccm_vlan = CFM_RANDOM_VLAN;
4542 s.ccm_vlan = atoi(cfm_ccm_vlan);
4543 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4548 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4550 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4552 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4553 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4555 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4558 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4559 * instead of obtaining it from the database. */
4561 iface_is_synthetic(const struct iface *iface)
4563 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4567 iface_validate_ofport__(size_t n, int64_t *ofport)
4569 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4570 ? u16_to_ofp(*ofport)
4575 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4577 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4581 iface_pick_ofport(const struct ovsrec_interface *cfg)
4583 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4584 return (requested_ofport != OFPP_NONE
4586 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4589 /* Port mirroring. */
4591 static struct mirror *
4592 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4596 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4597 if (uuid_equals(uuid, &m->uuid)) {
4605 bridge_configure_mirrors(struct bridge *br)
4607 const struct ovsdb_datum *mc;
4608 unsigned long *flood_vlans;
4609 struct mirror *m, *next;
4612 /* Get rid of deleted mirrors. */
4613 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4614 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4615 union ovsdb_atom atom;
4617 atom.uuid = m->uuid;
4618 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4623 /* Add new mirrors and reconfigure existing ones. */
4624 for (i = 0; i < br->cfg->n_mirrors; i++) {
4625 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4626 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4628 m = mirror_create(br, cfg);
4631 if (!mirror_configure(m)) {
4636 /* Update flooded vlans (for RSPAN). */
4637 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4638 br->cfg->n_flood_vlans);
4639 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4640 bitmap_free(flood_vlans);
4643 static struct mirror *
4644 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4648 m = xzalloc(sizeof *m);
4649 m->uuid = cfg->header_.uuid;
4650 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4652 m->name = xstrdup(cfg->name);
4658 mirror_destroy(struct mirror *m)
4661 struct bridge *br = m->bridge;
4664 ofproto_mirror_unregister(br->ofproto, m);
4667 hmap_remove(&br->mirrors, &m->hmap_node);
4674 mirror_collect_ports(struct mirror *m,
4675 struct ovsrec_port **in_ports, int n_in_ports,
4676 void ***out_portsp, size_t *n_out_portsp)
4678 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4679 size_t n_out_ports = 0;
4682 for (i = 0; i < n_in_ports; i++) {
4683 const char *name = in_ports[i]->name;
4684 struct port *port = port_lookup(m->bridge, name);
4686 out_ports[n_out_ports++] = port;
4688 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4689 "port %s", m->bridge->name, m->name, name);
4692 *out_portsp = out_ports;
4693 *n_out_portsp = n_out_ports;
4697 mirror_configure(struct mirror *m)
4699 const struct ovsrec_mirror *cfg = m->cfg;
4700 struct ofproto_mirror_settings s;
4703 if (strcmp(cfg->name, m->name)) {
4705 m->name = xstrdup(cfg->name);
4709 /* Get output port or VLAN. */
4710 if (cfg->output_port) {
4711 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4712 if (!s.out_bundle) {
4713 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4714 m->bridge->name, m->name);
4717 s.out_vlan = UINT16_MAX;
4719 if (cfg->output_vlan) {
4720 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4721 "output vlan; ignoring output vlan",
4722 m->bridge->name, m->name);
4724 } else if (cfg->output_vlan) {
4725 /* The database should prevent invalid VLAN values. */
4726 s.out_bundle = NULL;
4727 s.out_vlan = *cfg->output_vlan;
4729 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4730 m->bridge->name, m->name);
4734 /* Get port selection. */
4735 if (cfg->select_all) {
4736 size_t n_ports = hmap_count(&m->bridge->ports);
4737 void **ports = xmalloc(n_ports * sizeof *ports);
4742 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4752 /* Get ports, dropping ports that don't exist.
4753 * The IDL ensures that there are no duplicates. */
4754 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4755 &s.srcs, &s.n_srcs);
4756 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4757 &s.dsts, &s.n_dsts);
4760 /* Get VLAN selection. */
4761 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4764 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4767 if (s.srcs != s.dsts) {
4776 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4778 * This is deprecated. It is only for compatibility with broken device drivers
4779 * in old versions of Linux that do not properly support VLANs when VLAN
4780 * devices are not used. When broken device drivers are no longer in
4781 * widespread use, we will delete these interfaces. */
4783 static struct ovsrec_port **recs;
4784 static size_t n_recs, allocated_recs;
4786 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4787 * splinters are reconfigured. */
4789 register_rec(struct ovsrec_port *rec)
4791 if (n_recs >= allocated_recs) {
4792 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4794 recs[n_recs++] = rec;
4797 /* Frees all of the ports registered with register_reg(). */
4799 free_registered_recs(void)
4803 for (i = 0; i < n_recs; i++) {
4804 struct ovsrec_port *port = recs[i];
4807 for (j = 0; j < port->n_interfaces; j++) {
4808 struct ovsrec_interface *iface = port->interfaces[j];
4813 smap_destroy(&port->other_config);
4814 free(port->interfaces);
4822 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4825 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4827 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4831 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4834 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4835 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4836 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4839 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4840 * use, returns NULL.
4842 * Updates 'vlan_splinters_enabled_anywhere'. */
4843 static unsigned long int *
4844 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4846 unsigned long int *splinter_vlans;
4847 struct sset splinter_ifaces;
4848 const char *real_dev_name;
4849 struct shash *real_devs;
4850 struct shash_node *node;
4854 /* Free space allocated for synthesized ports and interfaces, since we're
4855 * in the process of reconstructing all of them. */
4856 free_registered_recs();
4858 splinter_vlans = bitmap_allocate(4096);
4859 sset_init(&splinter_ifaces);
4860 vlan_splinters_enabled_anywhere = false;
4861 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4862 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4865 for (j = 0; j < br_cfg->n_ports; j++) {
4866 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4869 for (k = 0; k < port_cfg->n_interfaces; k++) {
4870 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4872 if (vlan_splinters_is_enabled(iface_cfg)) {
4873 vlan_splinters_enabled_anywhere = true;
4874 sset_add(&splinter_ifaces, iface_cfg->name);
4875 vlan_bitmap_from_array__(port_cfg->trunks,
4881 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4882 bitmap_set1(splinter_vlans, *port_cfg->tag);
4887 if (!vlan_splinters_enabled_anywhere) {
4888 free(splinter_vlans);
4889 sset_destroy(&splinter_ifaces);
4893 HMAP_FOR_EACH (br, node, &all_bridges) {
4895 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4899 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4900 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4901 * device to be created for it. */
4902 bitmap_set0(splinter_vlans, 0);
4903 bitmap_set0(splinter_vlans, 4095);
4905 /* Delete all VLAN devices that we don't need. */
4907 real_devs = vlandev_get_real_devs();
4908 SHASH_FOR_EACH (node, real_devs) {
4909 const struct vlan_real_dev *real_dev = node->data;
4910 const struct vlan_dev *vlan_dev;
4911 bool real_dev_has_splinters;
4913 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4915 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4916 if (!real_dev_has_splinters
4917 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4918 struct netdev *netdev;
4920 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4921 if (!netdev_get_in4(netdev, NULL, NULL) ||
4922 !netdev_get_in6(netdev, NULL)) {
4923 /* It has an IP address configured, so we don't own
4924 * it. Don't delete it. */
4926 vlandev_del(vlan_dev->name);
4928 netdev_close(netdev);
4935 /* Add all VLAN devices that we need. */
4936 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4939 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4940 if (!vlandev_get_name(real_dev_name, vid)) {
4941 vlandev_add(real_dev_name, vid);
4948 sset_destroy(&splinter_ifaces);
4950 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4951 free(splinter_vlans);
4954 return splinter_vlans;
4957 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4960 configure_splinter_port(struct port *port)
4962 struct ofproto *ofproto = port->bridge->ofproto;
4963 ofp_port_t realdev_ofp_port;
4964 const char *realdev_name;
4965 struct iface *vlandev, *realdev;
4967 ofproto_bundle_unregister(port->bridge->ofproto, port);
4969 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4972 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4973 realdev = iface_lookup(port->bridge, realdev_name);
4974 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4976 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4980 static struct ovsrec_port *
4981 synthesize_splinter_port(const char *real_dev_name,
4982 const char *vlan_dev_name, int vid)
4984 struct ovsrec_interface *iface;
4985 struct ovsrec_port *port;
4987 iface = xmalloc(sizeof *iface);
4988 ovsrec_interface_init(iface);
4989 iface->name = xstrdup(vlan_dev_name);
4990 iface->type = "system";
4992 port = xmalloc(sizeof *port);
4993 ovsrec_port_init(port);
4994 port->interfaces = xmemdup(&iface, sizeof iface);
4995 port->n_interfaces = 1;
4996 port->name = xstrdup(vlan_dev_name);
4997 port->vlan_mode = "splinter";
4998 port->tag = xmalloc(sizeof *port->tag);
5001 smap_add(&port->other_config, "realdev", real_dev_name);
5007 /* For each interface with 'br' that has VLAN splinters enabled, adds a
5008 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
5009 * 1-bit in the 'splinter_vlans' bitmap. */
5011 add_vlan_splinter_ports(struct bridge *br,
5012 const unsigned long int *splinter_vlans,
5013 struct shash *ports)
5017 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
5018 * we're modifying 'ports'. */
5019 for (i = 0; i < br->cfg->n_ports; i++) {
5020 const char *name = br->cfg->ports[i]->name;
5021 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
5024 for (j = 0; j < port_cfg->n_interfaces; j++) {
5025 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
5027 if (vlan_splinters_is_enabled(iface_cfg)) {
5028 const char *real_dev_name;
5031 real_dev_name = iface_cfg->name;
5032 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
5033 const char *vlan_dev_name;
5035 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5037 && !shash_find(ports, vlan_dev_name)) {
5038 shash_add(ports, vlan_dev_name,
5039 synthesize_splinter_port(
5040 real_dev_name, vlan_dev_name, vid));
5049 mirror_refresh_stats(struct mirror *m)
5051 struct ofproto *ofproto = m->bridge->ofproto;
5052 uint64_t tx_packets, tx_bytes;
5053 const char *keys[2];
5055 size_t stat_cnt = 0;
5057 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5058 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5062 if (tx_packets != UINT64_MAX) {
5063 keys[stat_cnt] = "tx_packets";
5064 values[stat_cnt] = tx_packets;
5067 if (tx_bytes != UINT64_MAX) {
5068 keys[stat_cnt] = "tx_bytes";
5069 values[stat_cnt] = tx_bytes;
5073 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5077 * Add registered netdev and dpif types to ovsdb to allow external
5078 * applications to query the capabilities of the Open vSwitch instance
5079 * running on the node.
5082 discover_types(const struct ovsrec_open_vswitch *cfg)
5086 /* Datapath types. */
5088 dp_enumerate_types(&types);
5089 const char **datapath_types = sset_array(&types);
5090 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5091 sset_count(&types));
5092 free(datapath_types);
5093 sset_destroy(&types);
5097 netdev_enumerate_types(&types);
5098 const char **iface_types = sset_array(&types);
5099 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5101 sset_destroy(&types);