1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 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.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "mcast-snooping.h"
38 #include "meta-flow.h"
40 #include "ofp-print.h"
43 #include "ofproto/bond.h"
44 #include "ofproto/ofproto.h"
46 #include "poll-loop.h"
51 #include "socket-util.h"
53 #include "stream-ssl.h"
55 #include "system-stats.h"
60 #include "lib/vswitch-idl.h"
61 #include "xenserver.h"
63 #include "sflow_api.h"
64 #include "vlan-bitmap.h"
66 VLOG_DEFINE_THIS_MODULE(bridge);
68 COVERAGE_DEFINE(bridge_reconfigure);
71 /* These members are always valid.
73 * They are immutable: they never change between iface_create() and
75 struct list port_elem; /* Element in struct port's "ifaces" list. */
76 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
77 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
78 struct port *port; /* Containing port. */
79 char *name; /* Host network device name. */
80 struct netdev *netdev; /* Network device. */
81 ofp_port_t ofp_port; /* OpenFlow port number. */
84 /* These members are valid only within bridge_reconfigure(). */
85 const char *type; /* Usually same as cfg->type. */
86 const struct ovsrec_interface *cfg;
90 struct uuid uuid; /* UUID of this "mirror" record in database. */
91 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
92 struct bridge *bridge;
94 const struct ovsrec_mirror *cfg;
98 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
99 struct bridge *bridge;
102 const struct ovsrec_port *cfg;
104 /* An ordinary bridge port has 1 interface.
105 * A bridge port for bonding has at least 2 interfaces. */
106 struct list ifaces; /* List of "struct iface"s. */
110 struct hmap_node node; /* In 'all_bridges'. */
111 char *name; /* User-specified arbitrary name. */
112 char *type; /* Datapath type. */
113 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
114 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
115 const struct ovsrec_bridge *cfg;
117 /* OpenFlow switch processing. */
118 struct ofproto *ofproto; /* OpenFlow switch. */
121 struct hmap ports; /* "struct port"s indexed by name. */
122 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
123 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
125 /* Port mirroring. */
126 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
128 /* Used during reconfiguration. */
129 struct shash wanted_ports;
131 /* Synthetic local port if necessary. */
132 struct ovsrec_port synth_local_port;
133 struct ovsrec_interface synth_local_iface;
134 struct ovsrec_interface *synth_local_ifacep;
137 /* All bridges, indexed by name. */
138 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
140 /* OVSDB IDL used to obtain configuration. */
141 static struct ovsdb_idl *idl;
143 /* We want to complete daemonization, fully detaching from our parent process,
144 * only after we have completed our initial configuration, committed our state
145 * to the database, and received confirmation back from the database server
146 * that it applied the commit. This allows our parent process to know that,
147 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
148 * to have already been filled in. (It is only "very likely" rather than
149 * certain because there is always a slim possibility that the transaction will
150 * fail or that some other client has added new bridges, ports, etc. while
151 * ovs-vswitchd was configuring using an old configuration.)
153 * We only need to do this once for our initial configuration at startup, so
154 * 'initial_config_done' tracks whether we've already done it. While we are
155 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
156 * itself and is otherwise NULL. */
157 static bool initial_config_done;
158 static struct ovsdb_idl_txn *daemonize_txn;
160 /* Most recently processed IDL sequence number. */
161 static unsigned int idl_seqno;
163 /* Track changes to port connectivity. */
164 static uint64_t connectivity_seqno = LLONG_MIN;
166 /* Status update to database.
168 * Some information in the database must be kept as up-to-date as possible to
169 * allow controllers to respond rapidly to network outages. Those status are
170 * updated via the 'status_txn'.
172 * We use the global connectivity sequence number to detect the status change.
173 * Also, to prevent the status update from sending too much to the database,
174 * we check the return status of each update transaction and do not start new
175 * update if the previous transaction status is 'TXN_INCOMPLETE'.
177 * 'statux_txn' is NULL if there is no ongoing status update.
179 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
180 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
181 * which will cause the main thread wake up soon and retry the status update.
183 static struct ovsdb_idl_txn *status_txn;
184 static bool status_txn_try_again;
186 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
187 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
188 #define STATUS_CHECK_AGAIN_MSEC 100
190 /* Each time this timer expires, the bridge fetches interface and mirror
191 * statistics and pushes them into the database. */
192 static int stats_timer_interval;
193 static long long int stats_timer = LLONG_MIN;
195 /* Current stats database transaction, NULL if there is no ongoing
197 static struct ovsdb_idl_txn *stats_txn;
199 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
200 * expensive. This can cause bridge_reconfigure() to take a long time during
201 * which no other work can be done. To deal with this problem, we limit port
202 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
203 * call to bridge_reconfigure(). If there is more work to do after the limit
204 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
205 * This allows the rest of the code to catch up on important things like
206 * forwarding packets. */
207 #define OFP_PORT_ACTION_WINDOW 10
209 static void add_del_bridges(const struct ovsrec_open_vswitch *);
210 static void bridge_run__(void);
211 static void bridge_create(const struct ovsrec_bridge *);
212 static void bridge_destroy(struct bridge *);
213 static struct bridge *bridge_lookup(const char *name);
214 static unixctl_cb_func bridge_unixctl_dump_flows;
215 static unixctl_cb_func bridge_unixctl_reconnect;
216 static size_t bridge_get_controllers(const struct bridge *br,
217 struct ovsrec_controller ***controllersp);
218 static void bridge_collect_wanted_ports(struct bridge *,
219 const unsigned long *splinter_vlans,
220 struct shash *wanted_ports);
221 static void bridge_delete_ofprotos(void);
222 static void bridge_delete_or_reconfigure_ports(struct bridge *);
223 static void bridge_del_ports(struct bridge *,
224 const struct shash *wanted_ports);
225 static void bridge_add_ports(struct bridge *,
226 const struct shash *wanted_ports);
228 static void bridge_configure_datapath_id(struct bridge *);
229 static void bridge_configure_netflow(struct bridge *);
230 static void bridge_configure_forward_bpdu(struct bridge *);
231 static void bridge_configure_mac_table(struct bridge *);
232 static void bridge_configure_mcast_snooping(struct bridge *);
233 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
234 static void bridge_configure_ipfix(struct bridge *);
235 static void bridge_configure_stp(struct bridge *);
236 static void bridge_configure_rstp(struct bridge *);
237 static void bridge_configure_tables(struct bridge *);
238 static void bridge_configure_dp_desc(struct bridge *);
239 static void bridge_configure_remotes(struct bridge *,
240 const struct sockaddr_in *managers,
242 static void bridge_pick_local_hw_addr(struct bridge *,
243 uint8_t ea[ETH_ADDR_LEN],
244 struct iface **hw_addr_iface);
245 static uint64_t bridge_pick_datapath_id(struct bridge *,
246 const uint8_t bridge_ea[ETH_ADDR_LEN],
247 struct iface *hw_addr_iface);
248 static uint64_t dpid_from_hash(const void *, size_t nbytes);
249 static bool bridge_has_bond_fake_iface(const struct bridge *,
251 static bool port_is_bond_fake_iface(const struct port *);
253 static unixctl_cb_func qos_unixctl_show;
255 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
256 static void port_del_ifaces(struct port *);
257 static void port_destroy(struct port *);
258 static struct port *port_lookup(const struct bridge *, const char *name);
259 static void port_configure(struct port *);
260 static struct lacp_settings *port_configure_lacp(struct port *,
261 struct lacp_settings *);
262 static void port_configure_bond(struct port *, struct bond_settings *);
263 static bool port_is_synthetic(const struct port *);
265 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
266 static void run_system_stats(void);
268 static void bridge_configure_mirrors(struct bridge *);
269 static struct mirror *mirror_create(struct bridge *,
270 const struct ovsrec_mirror *);
271 static void mirror_destroy(struct mirror *);
272 static bool mirror_configure(struct mirror *);
273 static void mirror_refresh_stats(struct mirror *);
275 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
276 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
277 const struct ovsrec_port *);
278 static bool iface_is_internal(const struct ovsrec_interface *iface,
279 const struct ovsrec_bridge *br);
280 static const char *iface_get_type(const struct ovsrec_interface *,
281 const struct ovsrec_bridge *);
282 static void iface_destroy(struct iface *);
283 static void iface_destroy__(struct iface *);
284 static struct iface *iface_lookup(const struct bridge *, const char *name);
285 static struct iface *iface_find(const char *name);
286 static struct iface *iface_from_ofp_port(const struct bridge *,
287 ofp_port_t ofp_port);
288 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
289 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
290 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
291 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
292 static void iface_configure_cfm(struct iface *);
293 static void iface_refresh_cfm_stats(struct iface *);
294 static void iface_refresh_stats(struct iface *);
295 static void iface_refresh_netdev_status(struct iface *);
296 static void iface_refresh_ofproto_status(struct iface *);
297 static bool iface_is_synthetic(const struct iface *);
298 static ofp_port_t iface_get_requested_ofp_port(
299 const struct ovsrec_interface *);
300 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
302 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
304 * This is deprecated. It is only for compatibility with broken device drivers
305 * in old versions of Linux that do not properly support VLANs when VLAN
306 * devices are not used. When broken device drivers are no longer in
307 * widespread use, we will delete these interfaces. */
309 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
310 static bool vlan_splinters_enabled_anywhere;
312 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
313 static unsigned long int *collect_splinter_vlans(
314 const struct ovsrec_open_vswitch *);
315 static void configure_splinter_port(struct port *);
316 static void add_vlan_splinter_ports(struct bridge *,
317 const unsigned long int *splinter_vlans,
318 struct shash *ports);
321 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
323 struct shash iface_hints;
324 static bool initialized = false;
331 shash_init(&iface_hints);
334 for (i = 0; i < cfg->n_bridges; i++) {
335 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
338 for (j = 0; j < br_cfg->n_ports; j++) {
339 struct ovsrec_port *port_cfg = br_cfg->ports[j];
342 for (k = 0; k < port_cfg->n_interfaces; k++) {
343 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
344 struct iface_hint *iface_hint;
346 iface_hint = xmalloc(sizeof *iface_hint);
347 iface_hint->br_name = br_cfg->name;
348 iface_hint->br_type = br_cfg->datapath_type;
349 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
351 shash_add(&iface_hints, if_cfg->name, iface_hint);
357 ofproto_init(&iface_hints);
359 shash_destroy_free_data(&iface_hints);
363 /* Public functions. */
365 /* Initializes the bridge module, configuring it to obtain its configuration
366 * from an OVSDB server accessed over 'remote', which should be a string in a
367 * form acceptable to ovsdb_idl_create(). */
369 bridge_init(const char *remote)
371 /* Create connection to database. */
372 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
373 idl_seqno = ovsdb_idl_get_seqno(idl);
374 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
375 ovsdb_idl_verify_write_only(idl);
377 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
378 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
379 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
380 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
381 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
382 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
383 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
385 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
386 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
387 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
388 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
389 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
390 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
392 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
393 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
394 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
395 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
396 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
408 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
409 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
410 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
411 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
412 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
413 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
414 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
415 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
416 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
417 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
418 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
420 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
421 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
422 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
423 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
425 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
427 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
429 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
430 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
432 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
433 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
434 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
435 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
437 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
438 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
439 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
440 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
441 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
443 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
445 /* Register unixctl commands. */
446 unixctl_command_register("qos/show", "interface", 1, 1,
447 qos_unixctl_show, NULL);
448 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
449 bridge_unixctl_dump_flows, NULL);
450 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
451 bridge_unixctl_reconnect, NULL);
463 struct bridge *br, *next_br;
465 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
468 ovsdb_idl_destroy(idl);
471 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
472 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
473 * responsible for freeing '*managersp' (with free()).
475 * You may be asking yourself "why does ovs-vswitchd care?", because
476 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
477 * should not be and in fact is not directly involved in that. But
478 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
479 * it has to tell in-band control where the managers are to enable that.
480 * (Thus, only managers connected in-band are collected.)
483 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
484 struct sockaddr_in **managersp, size_t *n_managersp)
486 struct sockaddr_in *managers = NULL;
487 size_t n_managers = 0;
491 /* Collect all of the potential targets from the "targets" columns of the
492 * rows pointed to by "manager_options", excluding any that are
495 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
496 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
498 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
499 sset_find_and_delete(&targets, m->target);
501 sset_add(&targets, m->target);
505 /* Now extract the targets' IP addresses. */
506 if (!sset_is_empty(&targets)) {
509 managers = xmalloc(sset_count(&targets) * sizeof *managers);
510 SSET_FOR_EACH (target, &targets) {
512 struct sockaddr_storage ss;
513 struct sockaddr_in in;
516 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
518 && sa.ss.ss_family == AF_INET) {
519 managers[n_managers++] = sa.in;
523 sset_destroy(&targets);
525 *managersp = managers;
526 *n_managersp = n_managers;
530 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
532 unsigned long int *splinter_vlans;
533 struct sockaddr_in *managers;
534 struct bridge *br, *next;
535 int sflow_bridge_number;
538 COVERAGE_INC(bridge_reconfigure);
540 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
541 OFPROTO_FLOW_LIMIT_DEFAULT));
542 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
543 OFPROTO_MAX_IDLE_DEFAULT));
544 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
546 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
549 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
550 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
552 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
553 * to 'ovs_cfg', with only very minimal configuration otherwise.
555 * This is mostly an update to bridge data structures. Nothing is pushed
556 * down to ofproto or lower layers. */
557 add_del_bridges(ovs_cfg);
558 splinter_vlans = collect_splinter_vlans(ovs_cfg);
559 HMAP_FOR_EACH (br, node, &all_bridges) {
560 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
561 bridge_del_ports(br, &br->wanted_ports);
563 free(splinter_vlans);
565 /* Start pushing configuration changes down to the ofproto layer:
567 * - Delete ofprotos that are no longer configured.
569 * - Delete ports that are no longer configured.
571 * - Reconfigure existing ports to their desired configurations, or
572 * delete them if not possible.
574 * We have to do all the deletions before we can do any additions, because
575 * the ports to be added might require resources that will be freed up by
576 * deletions (they might especially overlap in name). */
577 bridge_delete_ofprotos();
578 HMAP_FOR_EACH (br, node, &all_bridges) {
580 bridge_delete_or_reconfigure_ports(br);
584 /* Finish pushing configuration changes to the ofproto layer:
586 * - Create ofprotos that are missing.
588 * - Add ports that are missing. */
589 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
593 error = ofproto_create(br->name, br->type, &br->ofproto);
595 VLOG_ERR("failed to create bridge %s: %s", br->name,
596 ovs_strerror(error));
597 shash_destroy(&br->wanted_ports);
602 HMAP_FOR_EACH (br, node, &all_bridges) {
603 bridge_add_ports(br, &br->wanted_ports);
604 shash_destroy(&br->wanted_ports);
607 reconfigure_system_stats(ovs_cfg);
609 /* Complete the configuration. */
610 sflow_bridge_number = 0;
611 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
612 HMAP_FOR_EACH (br, node, &all_bridges) {
615 /* We need the datapath ID early to allow LACP ports to use it as the
616 * default system ID. */
617 bridge_configure_datapath_id(br);
619 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
622 port_configure(port);
624 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
625 iface_set_ofport(iface->cfg, iface->ofp_port);
626 /* Clear eventual previous errors */
627 ovsrec_interface_set_error(iface->cfg, NULL);
628 iface_configure_cfm(iface);
629 iface_configure_qos(iface, port->cfg->qos);
630 iface_set_mac(br, port, iface);
631 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
635 bridge_configure_mirrors(br);
636 bridge_configure_forward_bpdu(br);
637 bridge_configure_mac_table(br);
638 bridge_configure_mcast_snooping(br);
639 bridge_configure_remotes(br, managers, n_managers);
640 bridge_configure_netflow(br);
641 bridge_configure_sflow(br, &sflow_bridge_number);
642 bridge_configure_ipfix(br);
643 bridge_configure_stp(br);
644 bridge_configure_rstp(br);
645 bridge_configure_tables(br);
646 bridge_configure_dp_desc(br);
650 /* The ofproto-dpif provider does some final reconfiguration in its
651 * ->type_run() function. We have to call it before notifying the database
652 * client that reconfiguration is complete, otherwise there is a very
653 * narrow race window in which e.g. ofproto/trace will not recognize the
654 * new configuration (sometimes this causes unit test failures). */
658 /* Delete ofprotos which aren't configured or have the wrong type. Create
659 * ofprotos which don't exist but need to. */
661 bridge_delete_ofprotos(void)
668 /* Delete ofprotos with no bridge or with the wrong type. */
671 ofproto_enumerate_types(&types);
672 SSET_FOR_EACH (type, &types) {
675 ofproto_enumerate_names(type, &names);
676 SSET_FOR_EACH (name, &names) {
677 br = bridge_lookup(name);
678 if (!br || strcmp(type, br->type)) {
679 ofproto_delete(name, type);
683 sset_destroy(&names);
684 sset_destroy(&types);
688 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
690 if (*n >= *allocated) {
691 ports = x2nrealloc(ports, allocated, sizeof *ports);
693 ports[(*n)++] = port;
698 bridge_delete_or_reconfigure_ports(struct bridge *br)
700 struct ofproto_port ofproto_port;
701 struct ofproto_port_dump dump;
703 struct sset ofproto_ports;
704 struct port *port, *port_next;
706 /* List of "ofp_port"s to delete. We make a list instead of deleting them
707 * right away because ofproto implementations aren't necessarily able to
708 * iterate through a changing list of ports in an entirely robust way. */
715 sset_init(&ofproto_ports);
717 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
718 * that are not configured in the database. (This commonly happens when
719 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
721 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
722 * that have the wrong OpenFlow port number (and arrange to add them back
723 * with the correct OpenFlow port number). */
724 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
725 ofp_port_t requested_ofp_port;
728 sset_add(&ofproto_ports, ofproto_port.name);
730 iface = iface_lookup(br, ofproto_port.name);
732 /* No such iface is configured, so we should delete this
735 * As a corner case exception, keep the port if it's a bond fake
737 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
738 && !strcmp(ofproto_port.type, "internal")) {
744 if (strcmp(ofproto_port.type, iface->type)
745 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
746 /* The interface is the wrong type or can't be configured.
751 /* If the requested OpenFlow port for 'iface' changed, and it's not
752 * already the correct port, then we might want to temporarily delete
753 * this interface, so we can add it back again with the new OpenFlow
755 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
756 if (iface->ofp_port != OFPP_LOCAL &&
757 requested_ofp_port != OFPP_NONE &&
758 requested_ofp_port != iface->ofp_port) {
759 ofp_port_t victim_request;
760 struct iface *victim;
762 /* Check for an existing OpenFlow port currently occupying
763 * 'iface''s requested port number. If there isn't one, then
764 * delete this port. Otherwise we need to consider further. */
765 victim = iface_from_ofp_port(br, requested_ofp_port);
770 /* 'victim' is a port currently using 'iface''s requested port
771 * number. Unless 'victim' specifically requested that port
772 * number, too, then we can delete both 'iface' and 'victim'
773 * temporarily. (We'll add both of them back again later with new
774 * OpenFlow port numbers.)
776 * If 'victim' did request port number 'requested_ofp_port', just
777 * like 'iface', then that's a configuration inconsistency that we
778 * can't resolve. We might as well let it keep its current port
780 victim_request = iface_get_requested_ofp_port(victim->cfg);
781 if (victim_request != requested_ofp_port) {
782 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
783 iface_destroy(victim);
792 iface_destroy(iface);
793 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
795 for (i = 0; i < n; i++) {
796 ofproto_port_del(br->ofproto, del[i]);
800 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
801 * that we didn't see when we iterated through the datapath, i.e. ports
802 * that disappeared underneath use. This is an unusual situation, but it
803 * can happen in some cases:
805 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
806 * idea but could happen).
808 * - The port represented a device that disappeared, e.g. a tuntap
809 * device destroyed via "tunctl -d", a physical Ethernet device
810 * whose module was just unloaded via "rmmod", or a virtual NIC for a
811 * VM whose VM was just terminated. */
812 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
813 struct iface *iface, *iface_next;
815 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
816 if (!sset_contains(&ofproto_ports, iface->name)) {
817 iface_destroy__(iface);
821 if (list_is_empty(&port->ifaces)) {
825 sset_destroy(&ofproto_ports);
829 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
830 bool with_requested_port)
832 struct shash_node *port_node;
834 SHASH_FOR_EACH (port_node, wanted_ports) {
835 const struct ovsrec_port *port_cfg = port_node->data;
838 for (i = 0; i < port_cfg->n_interfaces; i++) {
839 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
840 ofp_port_t requested_ofp_port;
842 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
843 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
844 struct iface *iface = iface_lookup(br, iface_cfg->name);
847 iface_create(br, iface_cfg, port_cfg);
855 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
857 /* First add interfaces that request a particular port number. */
858 bridge_add_ports__(br, wanted_ports, true);
860 /* Then add interfaces that want automatic port number assignment.
861 * We add these afterward to avoid accidentally taking a specifically
862 * requested port number. */
863 bridge_add_ports__(br, wanted_ports, false);
867 port_configure(struct port *port)
869 const struct ovsrec_port *cfg = port->cfg;
870 struct bond_settings bond_settings;
871 struct lacp_settings lacp_settings;
872 struct ofproto_bundle_settings s;
875 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
876 configure_splinter_port(port);
885 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
886 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
887 s.slaves[s.n_slaves++] = iface->ofp_port;
892 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
896 /* Get VLAN trunks. */
899 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
903 if (cfg->vlan_mode) {
904 if (!strcmp(cfg->vlan_mode, "access")) {
905 s.vlan_mode = PORT_VLAN_ACCESS;
906 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
907 s.vlan_mode = PORT_VLAN_TRUNK;
908 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
909 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
910 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
911 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
913 /* This "can't happen" because ovsdb-server should prevent it. */
914 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
915 "back to trunk mode", port->name, cfg->vlan_mode);
916 s.vlan_mode = PORT_VLAN_TRUNK;
920 s.vlan_mode = PORT_VLAN_ACCESS;
922 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
926 s.vlan_mode = PORT_VLAN_TRUNK;
929 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
932 /* Get LACP settings. */
933 s.lacp = port_configure_lacp(port, &lacp_settings);
937 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
938 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
939 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
942 s.lacp_slaves = NULL;
945 /* Get bond settings. */
946 if (s.n_slaves > 1) {
947 s.bond = &bond_settings;
948 port_configure_bond(port, &bond_settings);
951 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
952 netdev_set_miimon_interval(iface->netdev, 0);
957 ofproto_bundle_register(port->bridge->ofproto, port, &s);
965 /* Pick local port hardware address and datapath ID for 'br'. */
967 bridge_configure_datapath_id(struct bridge *br)
969 uint8_t ea[ETH_ADDR_LEN];
971 struct iface *local_iface;
972 struct iface *hw_addr_iface;
975 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
976 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
978 int error = netdev_set_etheraddr(local_iface->netdev, ea);
980 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
981 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
982 "Ethernet address: %s",
983 br->name, ovs_strerror(error));
986 memcpy(br->ea, ea, ETH_ADDR_LEN);
988 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
989 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
990 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
991 ofproto_set_datapath_id(br->ofproto, dpid);
994 dpid_string = xasprintf("%016"PRIx64, dpid);
995 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
999 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1002 bridge_get_allowed_versions(struct bridge *br)
1004 if (!br->cfg->n_protocols)
1007 return ofputil_versions_from_strings(br->cfg->protocols,
1008 br->cfg->n_protocols);
1011 /* Set NetFlow configuration on 'br'. */
1013 bridge_configure_netflow(struct bridge *br)
1015 struct ovsrec_netflow *cfg = br->cfg->netflow;
1016 struct netflow_options opts;
1019 ofproto_set_netflow(br->ofproto, NULL);
1023 memset(&opts, 0, sizeof opts);
1025 /* Get default NetFlow configuration from datapath.
1026 * Apply overrides from 'cfg'. */
1027 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1028 if (cfg->engine_type) {
1029 opts.engine_type = *cfg->engine_type;
1031 if (cfg->engine_id) {
1032 opts.engine_id = *cfg->engine_id;
1035 /* Configure active timeout interval. */
1036 opts.active_timeout = cfg->active_timeout;
1037 if (!opts.active_timeout) {
1038 opts.active_timeout = -1;
1039 } else if (opts.active_timeout < 0) {
1040 VLOG_WARN("bridge %s: active timeout interval set to negative "
1041 "value, using default instead (%d seconds)", br->name,
1042 NF_ACTIVE_TIMEOUT_DEFAULT);
1043 opts.active_timeout = -1;
1046 /* Add engine ID to interface number to disambiguate bridgs? */
1047 opts.add_id_to_iface = cfg->add_id_to_interface;
1048 if (opts.add_id_to_iface) {
1049 if (opts.engine_id > 0x7f) {
1050 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1051 "another vswitch, choose an engine id less than 128",
1054 if (hmap_count(&br->ports) > 508) {
1055 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1056 "another port when more than 508 ports are used",
1062 sset_init(&opts.collectors);
1063 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1066 if (ofproto_set_netflow(br->ofproto, &opts)) {
1067 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1069 sset_destroy(&opts.collectors);
1072 /* Set sFlow configuration on 'br'. */
1074 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1076 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1077 struct ovsrec_controller **controllers;
1078 struct ofproto_sflow_options oso;
1079 size_t n_controllers;
1083 ofproto_set_sflow(br->ofproto, NULL);
1087 memset(&oso, 0, sizeof oso);
1089 sset_init(&oso.targets);
1090 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1092 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1093 if (cfg->sampling) {
1094 oso.sampling_rate = *cfg->sampling;
1097 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1099 oso.polling_interval = *cfg->polling;
1102 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1104 oso.header_len = *cfg->header;
1107 oso.sub_id = (*sflow_bridge_number)++;
1108 oso.agent_device = cfg->agent;
1110 oso.control_ip = NULL;
1111 n_controllers = bridge_get_controllers(br, &controllers);
1112 for (i = 0; i < n_controllers; i++) {
1113 if (controllers[i]->local_ip) {
1114 oso.control_ip = controllers[i]->local_ip;
1118 ofproto_set_sflow(br->ofproto, &oso);
1120 sset_destroy(&oso.targets);
1123 /* Returns whether a IPFIX row is valid. */
1125 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1127 return ipfix && ipfix->n_targets > 0;
1130 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1132 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1133 const struct bridge *br)
1135 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1138 /* Set IPFIX configuration on 'br'. */
1140 bridge_configure_ipfix(struct bridge *br)
1142 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1143 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1144 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1145 struct ofproto_ipfix_bridge_exporter_options be_opts;
1146 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1147 size_t n_fe_opts = 0;
1149 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1150 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1155 if (!valid_be_cfg && n_fe_opts == 0) {
1156 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1161 memset(&be_opts, 0, sizeof be_opts);
1163 sset_init(&be_opts.targets);
1164 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1166 if (be_cfg->sampling) {
1167 be_opts.sampling_rate = *be_cfg->sampling;
1169 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1171 if (be_cfg->obs_domain_id) {
1172 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1174 if (be_cfg->obs_point_id) {
1175 be_opts.obs_point_id = *be_cfg->obs_point_id;
1177 if (be_cfg->cache_active_timeout) {
1178 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1180 if (be_cfg->cache_max_flows) {
1181 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1184 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1185 "enable-tunnel-sampling", true);
1187 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1188 "enable-input-sampling", false);
1190 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1191 "enable-output-sampling", false);
1194 if (n_fe_opts > 0) {
1195 struct ofproto_ipfix_flow_exporter_options *opts;
1196 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1198 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1199 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1200 opts->collector_set_id = fe_cfg->id;
1201 sset_init(&opts->targets);
1202 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1203 fe_cfg->ipfix->n_targets);
1204 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1205 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1206 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1207 ? *fe_cfg->ipfix->cache_max_flows : 0;
1213 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1217 sset_destroy(&be_opts.targets);
1220 if (n_fe_opts > 0) {
1221 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1223 for (i = 0; i < n_fe_opts; i++) {
1224 sset_destroy(&opts->targets);
1232 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1233 struct ofproto_port_stp_settings *port_s,
1234 int *port_num_counter, unsigned long *port_num_bitmap)
1236 const char *config_str;
1237 struct iface *iface;
1239 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1240 port_s->enable = false;
1243 port_s->enable = true;
1246 /* STP over bonds is not supported. */
1247 if (!list_is_singleton(&port->ifaces)) {
1248 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1250 port_s->enable = false;
1254 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1256 /* Internal ports shouldn't participate in spanning tree, so
1258 if (!strcmp(iface->type, "internal")) {
1259 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1260 port_s->enable = false;
1264 /* STP on mirror output ports is not supported. */
1265 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1266 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1267 port_s->enable = false;
1271 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1273 unsigned long int port_num = strtoul(config_str, NULL, 0);
1274 int port_idx = port_num - 1;
1276 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1277 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1278 port_s->enable = false;
1282 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1283 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1284 port->name, port_num);
1285 port_s->enable = false;
1288 bitmap_set1(port_num_bitmap, port_idx);
1289 port_s->port_num = port_idx;
1291 if (*port_num_counter >= STP_MAX_PORTS) {
1292 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1293 port_s->enable = false;
1297 port_s->port_num = (*port_num_counter)++;
1300 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1302 port_s->path_cost = strtoul(config_str, NULL, 10);
1304 enum netdev_features current;
1307 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1308 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1309 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1312 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1314 port_s->priority = strtoul(config_str, NULL, 0);
1316 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1321 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1322 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1324 const char *config_str;
1325 struct iface *iface;
1327 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1328 port_s->enable = false;
1331 port_s->enable = true;
1334 /* RSTP over bonds is not supported. */
1335 if (!list_is_singleton(&port->ifaces)) {
1336 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1338 port_s->enable = false;
1342 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1344 /* Internal ports shouldn't participate in spanning tree, so
1346 if (!strcmp(iface->type, "internal")) {
1347 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1348 port_s->enable = false;
1352 /* RSTP on mirror output ports is not supported. */
1353 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1354 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1355 port_s->enable = false;
1359 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1361 unsigned long int port_num = strtoul(config_str, NULL, 0);
1362 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1363 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1364 port_s->enable = false;
1367 port_s->port_num = port_num;
1369 if (*port_num_counter >= RSTP_MAX_PORTS) {
1370 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1371 port_s->enable = false;
1374 /* If rstp-port-num is not specified, use 0.
1375 * rstp_port_set_port_number() will look for the first free one. */
1376 port_s->port_num = 0;
1379 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1381 port_s->path_cost = strtoul(config_str, NULL, 10);
1383 enum netdev_features current;
1386 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1387 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1388 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1391 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1393 port_s->priority = strtoul(config_str, NULL, 0);
1395 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1398 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1399 "rstp-port-admin-edge", false);
1400 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1401 "rstp-port-auto-edge", true);
1402 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1403 "rstp-port-mcheck", false);
1406 /* Set spanning tree configuration on 'br'. */
1408 bridge_configure_stp(struct bridge *br)
1410 struct ofproto_rstp_status rstp_status;
1412 ofproto_get_rstp_status(br->ofproto, &rstp_status);
1413 if (!br->cfg->stp_enable) {
1414 ofproto_set_stp(br->ofproto, NULL);
1415 } else if (rstp_status.enabled) {
1416 /* Do not activate STP if RSTP is enabled. */
1417 VLOG_ERR("STP cannot be enabled if RSTP is running.");
1418 ofproto_set_stp(br->ofproto, NULL);
1419 ovsrec_bridge_set_stp_enable(br->cfg, false);
1421 struct ofproto_stp_settings br_s;
1422 const char *config_str;
1424 int port_num_counter;
1425 unsigned long *port_num_bitmap;
1427 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1429 uint8_t ea[ETH_ADDR_LEN];
1431 if (eth_addr_from_string(config_str, ea)) {
1432 br_s.system_id = eth_addr_to_uint64(ea);
1434 br_s.system_id = eth_addr_to_uint64(br->ea);
1435 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1436 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1439 br_s.system_id = eth_addr_to_uint64(br->ea);
1442 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1444 br_s.priority = strtoul(config_str, NULL, 0);
1446 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1449 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1451 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1453 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1456 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1458 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1460 br_s.max_age = STP_DEFAULT_MAX_AGE;
1463 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1465 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1467 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1470 /* Configure STP on the bridge. */
1471 if (ofproto_set_stp(br->ofproto, &br_s)) {
1472 VLOG_ERR("bridge %s: could not enable STP", br->name);
1476 /* Users must either set the port number with the "stp-port-num"
1477 * configuration on all ports or none. If manual configuration
1478 * is not done, then we allocate them sequentially. */
1479 port_num_counter = 0;
1480 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1481 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1482 struct ofproto_port_stp_settings port_s;
1483 struct iface *iface;
1485 port_configure_stp(br->ofproto, port, &port_s,
1486 &port_num_counter, port_num_bitmap);
1488 /* As bonds are not supported, just apply configuration to
1489 * all interfaces. */
1490 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1491 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1493 VLOG_ERR("port %s: could not enable STP", port->name);
1499 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1500 && port_num_counter) {
1501 VLOG_ERR("bridge %s: must manually configure all STP port "
1502 "IDs or none, disabling", br->name);
1503 ofproto_set_stp(br->ofproto, NULL);
1505 bitmap_free(port_num_bitmap);
1510 bridge_configure_rstp(struct bridge *br)
1512 struct ofproto_stp_status stp_status;
1514 ofproto_get_stp_status(br->ofproto, &stp_status);
1515 if (!br->cfg->rstp_enable) {
1516 ofproto_set_rstp(br->ofproto, NULL);
1517 } else if (stp_status.enabled) {
1518 /* Do not activate RSTP if STP is enabled. */
1519 VLOG_ERR("RSTP cannot be enabled if STP is running.");
1520 ofproto_set_rstp(br->ofproto, NULL);
1521 ovsrec_bridge_set_rstp_enable(br->cfg, false);
1523 struct ofproto_rstp_settings br_s;
1524 const char *config_str;
1526 int port_num_counter;
1528 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1530 uint8_t ea[ETH_ADDR_LEN];
1532 if (eth_addr_from_string(config_str, ea)) {
1533 br_s.address = eth_addr_to_uint64(ea);
1536 br_s.address = eth_addr_to_uint64(br->ea);
1537 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1538 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1542 br_s.address = eth_addr_to_uint64(br->ea);
1545 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1547 br_s.priority = strtoul(config_str, NULL, 0);
1549 br_s.priority = RSTP_DEFAULT_PRIORITY;
1552 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1554 br_s.ageing_time = strtoul(config_str, NULL, 0);
1556 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1559 config_str = smap_get(&br->cfg->other_config,
1560 "rstp-force-protocol-version");
1562 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1564 br_s.force_protocol_version = FPV_DEFAULT;
1567 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1569 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1571 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1574 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1576 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1578 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1581 config_str = smap_get(&br->cfg->other_config,
1582 "rstp-transmit-hold-count");
1584 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1586 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1589 /* Configure RSTP on the bridge. */
1590 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1591 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1595 port_num_counter = 0;
1596 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1597 struct ofproto_port_rstp_settings port_s;
1598 struct iface *iface;
1600 port_configure_rstp(br->ofproto, port, &port_s,
1603 /* As bonds are not supported, just apply configuration to
1604 * all interfaces. */
1605 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1606 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1608 VLOG_ERR("port %s: could not enable RSTP", port->name);
1617 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1619 const struct port *port = port_lookup(br, name);
1620 return port && port_is_bond_fake_iface(port);
1624 port_is_bond_fake_iface(const struct port *port)
1626 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1630 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1632 struct bridge *br, *next;
1633 struct shash new_br;
1636 /* Collect new bridges' names and types. */
1637 shash_init(&new_br);
1638 for (i = 0; i < cfg->n_bridges; i++) {
1639 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1640 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1642 if (strchr(br_cfg->name, '/')) {
1643 /* Prevent remote ovsdb-server users from accessing arbitrary
1644 * directories, e.g. consider a bridge named "../../../etc/". */
1645 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1647 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1648 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1652 /* Get rid of deleted bridges or those whose types have changed.
1653 * Update 'cfg' of bridges that still exist. */
1654 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1655 br->cfg = shash_find_data(&new_br, br->name);
1656 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1657 br->cfg->datapath_type))) {
1662 /* Add new bridges. */
1663 for (i = 0; i < cfg->n_bridges; i++) {
1664 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1665 struct bridge *br = bridge_lookup(br_cfg->name);
1667 bridge_create(br_cfg);
1671 shash_destroy(&new_br);
1674 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1675 * Returns 0 if successful, otherwise a positive errno value. */
1677 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1678 struct netdev *netdev, char **errp)
1680 return netdev_set_config(netdev, &iface_cfg->options, errp);
1683 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1684 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1686 * If successful, returns 0 and stores the network device in '*netdevp'. On
1687 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1689 iface_do_create(const struct bridge *br,
1690 const struct ovsrec_interface *iface_cfg,
1691 const struct ovsrec_port *port_cfg,
1692 ofp_port_t *ofp_portp, struct netdev **netdevp,
1695 struct netdev *netdev = NULL;
1698 if (netdev_is_reserved_name(iface_cfg->name)) {
1699 VLOG_WARN("could not create interface %s, name is reserved",
1705 error = netdev_open(iface_cfg->name,
1706 iface_get_type(iface_cfg, br->cfg), &netdev);
1708 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1709 iface_cfg->name, ovs_strerror(error));
1713 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1718 *ofp_portp = iface_pick_ofport(iface_cfg);
1719 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1724 VLOG_INFO("bridge %s: added interface %s on port %d",
1725 br->name, iface_cfg->name, *ofp_portp);
1727 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1728 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1736 netdev_close(netdev);
1740 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1741 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1742 * automatically allocated for the iface. Takes ownership of and
1743 * deallocates 'if_cfg'.
1745 * Return true if an iface is successfully created, false otherwise. */
1747 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1748 const struct ovsrec_port *port_cfg)
1750 struct netdev *netdev;
1751 struct iface *iface;
1752 ofp_port_t ofp_port;
1757 /* Do the bits that can fail up front. */
1758 ovs_assert(!iface_lookup(br, iface_cfg->name));
1759 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1761 iface_clear_db_record(iface_cfg, errp);
1766 /* Get or create the port structure. */
1767 port = port_lookup(br, port_cfg->name);
1769 port = port_create(br, port_cfg);
1772 /* Create the iface structure. */
1773 iface = xzalloc(sizeof *iface);
1774 list_push_back(&port->ifaces, &iface->port_elem);
1775 hmap_insert(&br->iface_by_name, &iface->name_node,
1776 hash_string(iface_cfg->name, 0));
1778 iface->name = xstrdup(iface_cfg->name);
1779 iface->ofp_port = ofp_port;
1780 iface->netdev = netdev;
1781 iface->type = iface_get_type(iface_cfg, br->cfg);
1782 iface->cfg = iface_cfg;
1783 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1784 hash_ofp_port(ofp_port));
1786 /* Populate initial status in database. */
1787 iface_refresh_stats(iface);
1788 iface_refresh_netdev_status(iface);
1790 /* Add bond fake iface if necessary. */
1791 if (port_is_bond_fake_iface(port)) {
1792 struct ofproto_port ofproto_port;
1794 if (ofproto_port_query_by_name(br->ofproto, port->name,
1796 struct netdev *netdev;
1799 error = netdev_open(port->name, "internal", &netdev);
1801 ofp_port_t fake_ofp_port = OFPP_NONE;
1802 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1803 netdev_close(netdev);
1805 VLOG_WARN("could not open network device %s (%s)",
1806 port->name, ovs_strerror(error));
1809 /* Already exists, nothing to do. */
1810 ofproto_port_destroy(&ofproto_port);
1817 /* Set forward BPDU option. */
1819 bridge_configure_forward_bpdu(struct bridge *br)
1821 ofproto_set_forward_bpdu(br->ofproto,
1822 smap_get_bool(&br->cfg->other_config,
1827 /* Set MAC learning table configuration for 'br'. */
1829 bridge_configure_mac_table(struct bridge *br)
1831 const char *idle_time_str;
1834 const char *mac_table_size_str;
1837 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1838 idle_time = (idle_time_str && atoi(idle_time_str)
1839 ? atoi(idle_time_str)
1840 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1842 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1843 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1844 ? atoi(mac_table_size_str)
1847 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1850 /* Set multicast snooping table configuration for 'br'. */
1852 bridge_configure_mcast_snooping(struct bridge *br)
1854 if (!br->cfg->mcast_snooping_enable) {
1855 ofproto_set_mcast_snooping(br->ofproto, NULL);
1858 struct ofproto_mcast_snooping_settings br_s;
1859 const char *idle_time_str;
1860 const char *max_entries_str;
1862 idle_time_str = smap_get(&br->cfg->other_config,
1863 "mcast-snooping-aging-time");
1864 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1865 ? atoi(idle_time_str)
1866 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1868 max_entries_str = smap_get(&br->cfg->other_config,
1869 "mcast-snooping-table-size");
1870 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1871 ? atoi(max_entries_str)
1872 : MCAST_DEFAULT_MAX_ENTRIES);
1874 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1875 "mcast-snooping-disable-flood-unregistered",
1878 /* Configure multicast snooping on the bridge */
1879 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1880 VLOG_ERR("bridge %s: could not enable multicast snooping",
1885 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1886 bool flood = smap_get_bool(&port->cfg->other_config,
1887 "mcast-snooping-flood", false);
1888 if (ofproto_port_set_mcast_snooping(br->ofproto, port, flood)) {
1889 VLOG_ERR("port %s: could not configure mcast snooping",
1897 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1898 const struct port *fake_br, struct iface **hw_addr_iface)
1900 struct hmapx mirror_output_ports;
1902 bool found_addr = false;
1906 /* Mirror output ports don't participate in picking the local hardware
1907 * address. ofproto can't help us find out whether a given port is a
1908 * mirror output because we haven't configured mirrors yet, so we need to
1909 * accumulate them ourselves. */
1910 hmapx_init(&mirror_output_ports);
1911 for (i = 0; i < br->cfg->n_mirrors; i++) {
1912 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1913 if (m->output_port) {
1914 hmapx_add(&mirror_output_ports, m->output_port);
1918 /* Otherwise choose the minimum non-local MAC address among all of the
1920 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1921 uint8_t iface_ea[ETH_ADDR_LEN];
1922 struct iface *candidate;
1923 struct iface *iface;
1925 /* Mirror output ports don't participate. */
1926 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1930 /* Choose the MAC address to represent the port. */
1932 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1933 /* Find the interface with this Ethernet address (if any) so that
1934 * we can provide the correct devname to the caller. */
1935 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1936 uint8_t candidate_ea[ETH_ADDR_LEN];
1937 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1938 && eth_addr_equals(iface_ea, candidate_ea)) {
1943 /* Choose the interface whose MAC address will represent the port.
1944 * The Linux kernel bonding code always chooses the MAC address of
1945 * the first slave added to a bond, and the Fedora networking
1946 * scripts always add slaves to a bond in alphabetical order, so
1947 * for compatibility we choose the interface with the name that is
1948 * first in alphabetical order. */
1949 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1950 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1955 /* The local port doesn't count (since we're trying to choose its
1956 * MAC address anyway). */
1957 if (iface->ofp_port == OFPP_LOCAL) {
1961 /* For fake bridges we only choose from ports with the same tag */
1962 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1963 if (!port->cfg->tag) {
1966 if (*port->cfg->tag != *fake_br->cfg->tag) {
1972 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1978 /* Compare against our current choice. */
1979 if (!eth_addr_is_multicast(iface_ea) &&
1980 !eth_addr_is_local(iface_ea) &&
1981 !eth_addr_is_reserved(iface_ea) &&
1982 !eth_addr_is_zero(iface_ea) &&
1983 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1985 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1986 *hw_addr_iface = iface;
1992 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1993 *hw_addr_iface = NULL;
1996 hmapx_destroy(&mirror_output_ports);
2000 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2001 struct iface **hw_addr_iface)
2004 *hw_addr_iface = NULL;
2006 /* Did the user request a particular MAC? */
2007 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2008 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2009 if (eth_addr_is_multicast(ea)) {
2010 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2011 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2012 } else if (eth_addr_is_zero(ea)) {
2013 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2019 /* Find a local hw address */
2020 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2023 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2024 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2025 * an interface on 'br', then that interface must be passed in as
2026 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2027 * 'hw_addr_iface' must be passed in as a null pointer. */
2029 bridge_pick_datapath_id(struct bridge *br,
2030 const uint8_t bridge_ea[ETH_ADDR_LEN],
2031 struct iface *hw_addr_iface)
2034 * The procedure for choosing a bridge MAC address will, in the most
2035 * ordinary case, also choose a unique MAC that we can use as a datapath
2036 * ID. In some special cases, though, multiple bridges will end up with
2037 * the same MAC address. This is OK for the bridges, but it will confuse
2038 * the OpenFlow controller, because each datapath needs a unique datapath
2041 * Datapath IDs must be unique. It is also very desirable that they be
2042 * stable from one run to the next, so that policy set on a datapath
2045 const char *datapath_id;
2048 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2049 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2053 if (!hw_addr_iface) {
2055 * A purely internal bridge, that is, one that has no non-virtual
2056 * network devices on it at all, is difficult because it has no
2057 * natural unique identifier at all.
2059 * When the host is a XenServer, we handle this case by hashing the
2060 * host's UUID with the name of the bridge. Names of bridges are
2061 * persistent across XenServer reboots, although they can be reused if
2062 * an internal network is destroyed and then a new one is later
2063 * created, so this is fairly effective.
2065 * When the host is not a XenServer, we punt by using a random MAC
2066 * address on each run.
2068 const char *host_uuid = xenserver_get_host_uuid();
2070 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2071 dpid = dpid_from_hash(combined, strlen(combined));
2077 return eth_addr_to_uint64(bridge_ea);
2081 dpid_from_hash(const void *data, size_t n)
2083 uint8_t hash[SHA1_DIGEST_SIZE];
2085 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2086 sha1_bytes(data, n, hash);
2087 eth_addr_mark_random(hash);
2088 return eth_addr_to_uint64(hash);
2092 iface_refresh_netdev_status(struct iface *iface)
2096 enum netdev_features current;
2097 enum netdev_flags flags;
2098 const char *link_state;
2099 uint8_t mac[ETH_ADDR_LEN];
2100 int64_t bps, mtu_64, ifindex64, link_resets;
2103 if (iface_is_synthetic(iface)) {
2107 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2108 && !status_txn_try_again) {
2112 iface->change_seq = netdev_get_change_seq(iface->netdev);
2116 if (!netdev_get_status(iface->netdev, &smap)) {
2117 ovsrec_interface_set_status(iface->cfg, &smap);
2119 ovsrec_interface_set_status(iface->cfg, NULL);
2122 smap_destroy(&smap);
2124 error = netdev_get_flags(iface->netdev, &flags);
2126 const char *state = flags & NETDEV_UP ? "up" : "down";
2128 ovsrec_interface_set_admin_state(iface->cfg, state);
2130 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2133 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2134 ovsrec_interface_set_link_state(iface->cfg, link_state);
2136 link_resets = netdev_get_carrier_resets(iface->netdev);
2137 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2139 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2140 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2142 ovsrec_interface_set_duplex(iface->cfg,
2143 netdev_features_is_full_duplex(current)
2145 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2147 ovsrec_interface_set_duplex(iface->cfg, NULL);
2148 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2151 error = netdev_get_mtu(iface->netdev, &mtu);
2154 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2156 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2159 error = netdev_get_etheraddr(iface->netdev, mac);
2161 char mac_string[32];
2163 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2164 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2166 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2169 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2170 * if there is no valid ifindex number. */
2171 ifindex64 = netdev_get_ifindex(iface->netdev);
2172 if (ifindex64 < 0) {
2175 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2179 iface_refresh_ofproto_status(struct iface *iface)
2183 if (iface_is_synthetic(iface)) {
2187 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2191 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2193 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2196 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2198 || status_txn_try_again) {
2199 iface_refresh_cfm_stats(iface);
2202 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2204 || status_txn_try_again) {
2208 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2209 iface->ofp_port, &smap);
2210 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2211 smap_destroy(&smap);
2215 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2218 iface_refresh_cfm_stats(struct iface *iface)
2220 const struct ovsrec_interface *cfg = iface->cfg;
2221 struct cfm_status status;
2224 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2225 iface->ofp_port, &status);
2227 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2228 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2229 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2230 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2231 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2232 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2234 const char *reasons[CFM_FAULT_N_REASONS];
2235 int64_t cfm_health = status.health;
2236 int64_t cfm_flap_count = status.flap_count;
2237 bool faulted = status.faults != 0;
2240 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2243 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2244 int reason = 1 << i;
2245 if (status.faults & reason) {
2246 reasons[j++] = cfm_fault_reason_to_str(reason);
2249 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
2251 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2253 if (status.remote_opstate >= 0) {
2254 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2255 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2257 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2260 ovsrec_interface_set_cfm_remote_mpids(cfg,
2261 (const int64_t *)status.rmps,
2263 if (cfm_health >= 0) {
2264 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2266 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2274 iface_refresh_stats(struct iface *iface)
2276 #define IFACE_STATS \
2277 IFACE_STAT(rx_packets, "rx_packets") \
2278 IFACE_STAT(tx_packets, "tx_packets") \
2279 IFACE_STAT(rx_bytes, "rx_bytes") \
2280 IFACE_STAT(tx_bytes, "tx_bytes") \
2281 IFACE_STAT(rx_dropped, "rx_dropped") \
2282 IFACE_STAT(tx_dropped, "tx_dropped") \
2283 IFACE_STAT(rx_errors, "rx_errors") \
2284 IFACE_STAT(tx_errors, "tx_errors") \
2285 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2286 IFACE_STAT(rx_over_errors, "rx_over_err") \
2287 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2288 IFACE_STAT(collisions, "collisions")
2290 #define IFACE_STAT(MEMBER, NAME) + 1
2291 enum { N_IFACE_STATS = IFACE_STATS };
2293 int64_t values[N_IFACE_STATS];
2294 char *keys[N_IFACE_STATS];
2297 struct netdev_stats stats;
2299 if (iface_is_synthetic(iface)) {
2303 /* Intentionally ignore return value, since errors will set 'stats' to
2304 * all-1s, and we will deal with that correctly below. */
2305 netdev_get_stats(iface->netdev, &stats);
2307 /* Copy statistics into keys[] and values[]. */
2309 #define IFACE_STAT(MEMBER, NAME) \
2310 if (stats.MEMBER != UINT64_MAX) { \
2312 values[n] = stats.MEMBER; \
2317 ovs_assert(n <= N_IFACE_STATS);
2319 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2324 br_refresh_stp_status(struct bridge *br)
2326 struct smap smap = SMAP_INITIALIZER(&smap);
2327 struct ofproto *ofproto = br->ofproto;
2328 struct ofproto_stp_status status;
2330 if (ofproto_get_stp_status(ofproto, &status)) {
2334 if (!status.enabled) {
2335 ovsrec_bridge_set_status(br->cfg, NULL);
2339 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2340 STP_ID_ARGS(status.bridge_id));
2341 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2342 STP_ID_ARGS(status.designated_root));
2343 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2345 ovsrec_bridge_set_status(br->cfg, &smap);
2346 smap_destroy(&smap);
2350 port_refresh_stp_status(struct port *port)
2352 struct ofproto *ofproto = port->bridge->ofproto;
2353 struct iface *iface;
2354 struct ofproto_port_stp_status status;
2357 if (port_is_synthetic(port)) {
2361 /* STP doesn't currently support bonds. */
2362 if (!list_is_singleton(&port->ifaces)) {
2363 ovsrec_port_set_status(port->cfg, NULL);
2367 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2368 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2372 if (!status.enabled) {
2373 ovsrec_port_set_status(port->cfg, NULL);
2377 /* Set Status column. */
2379 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2380 smap_add(&smap, "stp_state", stp_state_name(status.state));
2381 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2382 smap_add(&smap, "stp_role", stp_role_name(status.role));
2383 ovsrec_port_set_status(port->cfg, &smap);
2384 smap_destroy(&smap);
2388 port_refresh_stp_stats(struct port *port)
2390 struct ofproto *ofproto = port->bridge->ofproto;
2391 struct iface *iface;
2392 struct ofproto_port_stp_stats stats;
2394 int64_t int_values[3];
2396 if (port_is_synthetic(port)) {
2400 /* STP doesn't currently support bonds. */
2401 if (!list_is_singleton(&port->ifaces)) {
2405 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2406 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2410 if (!stats.enabled) {
2411 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2415 /* Set Statistics column. */
2416 keys[0] = "stp_tx_count";
2417 int_values[0] = stats.tx_count;
2418 keys[1] = "stp_rx_count";
2419 int_values[1] = stats.rx_count;
2420 keys[2] = "stp_error_count";
2421 int_values[2] = stats.error_count;
2423 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2424 ARRAY_SIZE(int_values));
2428 br_refresh_rstp_status(struct bridge *br)
2430 struct smap smap = SMAP_INITIALIZER(&smap);
2431 struct ofproto *ofproto = br->ofproto;
2432 struct ofproto_rstp_status status;
2434 if (ofproto_get_rstp_status(ofproto, &status)) {
2437 if (!status.enabled) {
2438 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2441 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2442 RSTP_ID_ARGS(status.bridge_id));
2443 smap_add_format(&smap, "rstp_root_path_cost", "%d",
2444 status.root_path_cost);
2445 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2446 RSTP_ID_ARGS(status.root_id));
2447 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2448 RSTP_ID_ARGS(status.designated_id));
2449 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2450 status.designated_port_id);
2451 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2452 status.bridge_port_id);
2453 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2454 smap_destroy(&smap);
2458 port_refresh_rstp_status(struct port *port)
2460 struct ofproto *ofproto = port->bridge->ofproto;
2461 struct iface *iface;
2462 struct ofproto_port_rstp_status status;
2464 int64_t int_values[3];
2467 if (port_is_synthetic(port)) {
2471 /* RSTP doesn't currently support bonds. */
2472 if (!list_is_singleton(&port->ifaces)) {
2473 ovsrec_port_set_rstp_status(port->cfg, NULL);
2477 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2478 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2482 if (!status.enabled) {
2483 ovsrec_port_set_rstp_status(port->cfg, NULL);
2484 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2487 /* Set Status column. */
2490 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2492 smap_add_format(&smap, "rstp_port_role", "%s",
2493 rstp_port_role_name(status.role));
2494 smap_add_format(&smap, "rstp_port_state", "%s",
2495 rstp_state_name(status.state));
2497 ovsrec_port_set_rstp_status(port->cfg, &smap);
2498 smap_destroy(&smap);
2500 /* Set Statistics column. */
2501 keys[0] = "rstp_tx_count";
2502 int_values[0] = status.tx_count;
2503 keys[1] = "rstp_rx_count";
2504 int_values[1] = status.rx_count;
2505 keys[2] = "rstp_uptime";
2506 int_values[2] = status.uptime;
2507 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2508 ARRAY_SIZE(int_values));
2512 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2514 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2518 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2520 bool enable = enable_system_stats(cfg);
2522 system_stats_enable(enable);
2524 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2529 run_system_stats(void)
2531 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2534 stats = system_stats_run();
2536 struct ovsdb_idl_txn *txn;
2537 struct ovsdb_datum datum;
2539 txn = ovsdb_idl_txn_create(idl);
2540 ovsdb_datum_from_smap(&datum, stats);
2541 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2543 ovsdb_idl_txn_commit(txn);
2544 ovsdb_idl_txn_destroy(txn);
2551 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2554 case OFPCR12_ROLE_EQUAL:
2556 case OFPCR12_ROLE_MASTER:
2558 case OFPCR12_ROLE_SLAVE:
2560 case OFPCR12_ROLE_NOCHANGE:
2562 return "*** INVALID ROLE ***";
2567 refresh_controller_status(void)
2571 const struct ovsrec_controller *cfg;
2575 /* Accumulate status for controllers on all bridges. */
2576 HMAP_FOR_EACH (br, node, &all_bridges) {
2577 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2580 /* Update each controller in the database with current status. */
2581 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2582 struct ofproto_controller_info *cinfo =
2583 shash_find_data(&info, cfg->target);
2586 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2587 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2589 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2591 ovsrec_controller_set_is_connected(cfg, false);
2592 ovsrec_controller_set_role(cfg, NULL);
2593 ovsrec_controller_set_status(cfg, NULL);
2597 ofproto_free_ofproto_controller_info(&info);
2607 /* Let each datapath type do the work that it needs to do. */
2609 ofproto_enumerate_types(&types);
2610 SSET_FOR_EACH (type, &types) {
2611 ofproto_type_run(type);
2613 sset_destroy(&types);
2615 /* Let each bridge do the work that it needs to do. */
2616 HMAP_FOR_EACH (br, node, &all_bridges) {
2617 ofproto_run(br->ofproto);
2624 static struct ovsrec_open_vswitch null_cfg;
2625 const struct ovsrec_open_vswitch *cfg;
2627 bool vlan_splinters_changed;
2631 ovsrec_open_vswitch_init(&null_cfg);
2635 if (ovsdb_idl_is_lock_contended(idl)) {
2636 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2637 struct bridge *br, *next_br;
2639 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2640 "disabling this process (pid %ld) until it goes away",
2641 (long int) getpid());
2643 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2646 /* Since we will not be running system_stats_run() in this process
2647 * with the current situation of multiple ovs-vswitchd daemons,
2648 * disable system stats collection. */
2649 system_stats_enable(false);
2651 } else if (!ovsdb_idl_has_lock(idl)) {
2654 cfg = ovsrec_open_vswitch_first(idl);
2656 /* Initialize the ofproto library. This only needs to run once, but
2657 * it must be done after the configuration is set. If the
2658 * initialization has already occurred, bridge_init_ofproto()
2659 * returns immediately. */
2660 bridge_init_ofproto(cfg);
2662 /* Once the value of flow-restore-wait is false, we no longer should
2663 * check its value from the database. */
2664 if (cfg && ofproto_get_flow_restore_wait()) {
2665 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2666 "flow-restore-wait", false));
2671 /* Re-configure SSL. We do this on every trip through the main loop,
2672 * instead of just when the database changes, because the contents of the
2673 * key and certificate files can change without the database changing.
2675 * We do this before bridge_reconfigure() because that function might
2676 * initiate SSL connections and thus requires SSL to be configured. */
2677 if (cfg && cfg->ssl) {
2678 const struct ovsrec_ssl *ssl = cfg->ssl;
2680 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2681 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2684 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2685 * usage has changed. */
2686 vlan_splinters_changed = false;
2687 if (vlan_splinters_enabled_anywhere) {
2688 HMAP_FOR_EACH (br, node, &all_bridges) {
2689 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2690 vlan_splinters_changed = true;
2696 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2697 struct ovsdb_idl_txn *txn;
2699 idl_seqno = ovsdb_idl_get_seqno(idl);
2700 txn = ovsdb_idl_txn_create(idl);
2701 bridge_reconfigure(cfg ? cfg : &null_cfg);
2704 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2707 /* If we are completing our initial configuration for this run
2708 * of ovs-vswitchd, then keep the transaction around to monitor
2709 * it for completion. */
2710 if (initial_config_done) {
2711 /* Always sets the 'status_txn_try_again' to check again,
2712 * in case that this transaction fails. */
2713 status_txn_try_again = true;
2714 ovsdb_idl_txn_commit(txn);
2715 ovsdb_idl_txn_destroy(txn);
2717 initial_config_done = true;
2718 daemonize_txn = txn;
2722 if (daemonize_txn) {
2723 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2724 if (status != TXN_INCOMPLETE) {
2725 ovsdb_idl_txn_destroy(daemonize_txn);
2726 daemonize_txn = NULL;
2728 /* ovs-vswitchd has completed initialization, so allow the
2729 * process that forked us to exit successfully. */
2730 daemonize_complete();
2732 vlog_enable_async();
2734 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2738 /* Statistics update interval should always be greater than or equal to
2741 stats_interval = MAX(smap_get_int(&cfg->other_config,
2742 "stats-update-interval",
2745 stats_interval = 5000;
2747 if (stats_timer_interval != stats_interval) {
2748 stats_timer_interval = stats_interval;
2749 stats_timer = LLONG_MIN;
2752 /* Refresh interface and mirror stats if necessary. */
2753 if (time_msec() >= stats_timer) {
2754 enum ovsdb_idl_txn_status status;
2756 /* Rate limit the update. Do not start a new update if the
2757 * previous one is not done. */
2760 stats_txn = ovsdb_idl_txn_create(idl);
2761 HMAP_FOR_EACH (br, node, &all_bridges) {
2765 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2766 struct iface *iface;
2768 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2769 iface_refresh_stats(iface);
2771 port_refresh_stp_stats(port);
2773 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2774 mirror_refresh_stats(m);
2777 refresh_controller_status();
2781 status = ovsdb_idl_txn_commit(stats_txn);
2782 if (status != TXN_INCOMPLETE) {
2783 stats_timer = time_msec() + stats_timer_interval;
2784 ovsdb_idl_txn_destroy(stats_txn);
2792 /* Check the need to update status. */
2793 seq = seq_read(connectivity_seq_get());
2794 if (seq != connectivity_seqno || status_txn_try_again) {
2795 connectivity_seqno = seq;
2796 status_txn = ovsdb_idl_txn_create(idl);
2797 HMAP_FOR_EACH (br, node, &all_bridges) {
2800 br_refresh_stp_status(br);
2801 br_refresh_rstp_status(br);
2802 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2803 struct iface *iface;
2805 port_refresh_stp_status(port);
2806 port_refresh_rstp_status(port);
2807 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2808 iface_refresh_netdev_status(iface);
2809 iface_refresh_ofproto_status(iface);
2817 enum ovsdb_idl_txn_status status;
2819 status = ovsdb_idl_txn_commit(status_txn);
2820 /* Do not destroy "status_txn" if the transaction is
2821 * "TXN_INCOMPLETE". */
2822 if (status != TXN_INCOMPLETE) {
2823 ovsdb_idl_txn_destroy(status_txn);
2826 /* Sets the 'status_txn_try_again' if the transaction fails. */
2827 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2828 status_txn_try_again = false;
2830 status_txn_try_again = true;
2844 ovsdb_idl_wait(idl);
2845 if (daemonize_txn) {
2846 ovsdb_idl_txn_wait(daemonize_txn);
2850 ofproto_enumerate_types(&types);
2851 SSET_FOR_EACH (type, &types) {
2852 ofproto_type_wait(type);
2854 sset_destroy(&types);
2856 if (!hmap_is_empty(&all_bridges)) {
2859 HMAP_FOR_EACH (br, node, &all_bridges) {
2860 ofproto_wait(br->ofproto);
2863 poll_timer_wait_until(stats_timer);
2866 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2867 * transaction to complete. If the status update to database needs to be
2868 * run again (transaction fails), registers a timeout in
2869 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2870 * sequence number. */
2872 ovsdb_idl_txn_wait(status_txn);
2873 } else if (status_txn_try_again) {
2874 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2876 seq_wait(connectivity_seq_get(), connectivity_seqno);
2879 system_stats_wait();
2882 /* Adds some memory usage statistics for bridges into 'usage', for use with
2883 * memory_report(). */
2885 bridge_get_memory_usage(struct simap *usage)
2892 ofproto_enumerate_types(&types);
2893 SSET_FOR_EACH (type, &types) {
2894 ofproto_type_get_memory_usage(type, usage);
2896 sset_destroy(&types);
2898 HMAP_FOR_EACH (br, node, &all_bridges) {
2899 ofproto_get_memory_usage(br->ofproto, usage);
2903 /* QoS unixctl user interface functions. */
2905 struct qos_unixctl_show_cbdata {
2907 struct iface *iface;
2911 qos_unixctl_show_queue(unsigned int queue_id,
2912 const struct smap *details,
2913 struct iface *iface,
2916 struct netdev_queue_stats stats;
2917 struct smap_node *node;
2920 ds_put_cstr(ds, "\n");
2922 ds_put_format(ds, "Queue %u:\n", queue_id);
2924 ds_put_cstr(ds, "Default:\n");
2927 SMAP_FOR_EACH (node, details) {
2928 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2931 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2933 if (stats.tx_packets != UINT64_MAX) {
2934 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2937 if (stats.tx_bytes != UINT64_MAX) {
2938 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2941 if (stats.tx_errors != UINT64_MAX) {
2942 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2945 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2946 queue_id, ovs_strerror(error));
2951 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2952 const char *argv[], void *aux OVS_UNUSED)
2954 struct ds ds = DS_EMPTY_INITIALIZER;
2955 struct smap smap = SMAP_INITIALIZER(&smap);
2956 struct iface *iface;
2958 struct smap_node *node;
2960 iface = iface_find(argv[1]);
2962 unixctl_command_reply_error(conn, "no such interface");
2966 netdev_get_qos(iface->netdev, &type, &smap);
2968 if (*type != '\0') {
2969 struct netdev_queue_dump dump;
2970 struct smap details;
2971 unsigned int queue_id;
2973 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2975 SMAP_FOR_EACH (node, &smap) {
2976 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2979 smap_init(&details);
2980 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2981 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2983 smap_destroy(&details);
2985 unixctl_command_reply(conn, ds_cstr(&ds));
2987 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2988 unixctl_command_reply_error(conn, ds_cstr(&ds));
2991 smap_destroy(&smap);
2995 /* Bridge reconfiguration functions. */
2997 bridge_create(const struct ovsrec_bridge *br_cfg)
3001 ovs_assert(!bridge_lookup(br_cfg->name));
3002 br = xzalloc(sizeof *br);
3004 br->name = xstrdup(br_cfg->name);
3005 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3008 /* Derive the default Ethernet address from the bridge's UUID. This should
3009 * be unique and it will be stable between ovs-vswitchd runs. */
3010 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3011 eth_addr_mark_random(br->default_ea);
3013 hmap_init(&br->ports);
3014 hmap_init(&br->ifaces);
3015 hmap_init(&br->iface_by_name);
3016 hmap_init(&br->mirrors);
3018 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3022 bridge_destroy(struct bridge *br)
3025 struct mirror *mirror, *next_mirror;
3026 struct port *port, *next_port;
3028 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3031 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3032 mirror_destroy(mirror);
3035 hmap_remove(&all_bridges, &br->node);
3036 ofproto_destroy(br->ofproto);
3037 hmap_destroy(&br->ifaces);
3038 hmap_destroy(&br->ports);
3039 hmap_destroy(&br->iface_by_name);
3040 hmap_destroy(&br->mirrors);
3047 static struct bridge *
3048 bridge_lookup(const char *name)
3052 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3053 if (!strcmp(br->name, name)) {
3060 /* Handle requests for a listing of all flows known by the OpenFlow
3061 * stack, including those normally hidden. */
3063 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3064 const char *argv[], void *aux OVS_UNUSED)
3069 br = bridge_lookup(argv[1]);
3071 unixctl_command_reply_error(conn, "Unknown bridge");
3076 ofproto_get_all_flows(br->ofproto, &results);
3078 unixctl_command_reply(conn, ds_cstr(&results));
3079 ds_destroy(&results);
3082 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3083 * connections and reconnect. If BRIDGE is not specified, then all bridges
3084 * drop their controller connections and reconnect. */
3086 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3087 const char *argv[], void *aux OVS_UNUSED)
3091 br = bridge_lookup(argv[1]);
3093 unixctl_command_reply_error(conn, "Unknown bridge");
3096 ofproto_reconnect_controllers(br->ofproto);
3098 HMAP_FOR_EACH (br, node, &all_bridges) {
3099 ofproto_reconnect_controllers(br->ofproto);
3102 unixctl_command_reply(conn, NULL);
3106 bridge_get_controllers(const struct bridge *br,
3107 struct ovsrec_controller ***controllersp)
3109 struct ovsrec_controller **controllers;
3110 size_t n_controllers;
3112 controllers = br->cfg->controller;
3113 n_controllers = br->cfg->n_controller;
3115 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3121 *controllersp = controllers;
3123 return n_controllers;
3127 bridge_collect_wanted_ports(struct bridge *br,
3128 const unsigned long int *splinter_vlans,
3129 struct shash *wanted_ports)
3133 shash_init(wanted_ports);
3135 for (i = 0; i < br->cfg->n_ports; i++) {
3136 const char *name = br->cfg->ports[i]->name;
3137 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3138 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3142 if (bridge_get_controllers(br, NULL)
3143 && !shash_find(wanted_ports, br->name)) {
3144 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3145 br->name, br->name);
3147 ovsrec_interface_init(&br->synth_local_iface);
3148 ovsrec_port_init(&br->synth_local_port);
3150 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3151 br->synth_local_port.n_interfaces = 1;
3152 br->synth_local_port.name = br->name;
3154 br->synth_local_iface.name = br->name;
3155 br->synth_local_iface.type = "internal";
3157 br->synth_local_ifacep = &br->synth_local_iface;
3159 shash_add(wanted_ports, br->name, &br->synth_local_port);
3162 if (splinter_vlans) {
3163 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3167 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3168 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3169 * 'br' needs to complete its configuration. */
3171 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3173 struct shash_node *port_node;
3174 struct port *port, *next;
3176 /* Get rid of deleted ports.
3177 * Get rid of deleted interfaces on ports that still exist. */
3178 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3179 port->cfg = shash_find_data(wanted_ports, port->name);
3183 port_del_ifaces(port);
3187 /* Update iface->cfg and iface->type in interfaces that still exist. */
3188 SHASH_FOR_EACH (port_node, wanted_ports) {
3189 const struct ovsrec_port *port = port_node->data;
3192 for (i = 0; i < port->n_interfaces; i++) {
3193 const struct ovsrec_interface *cfg = port->interfaces[i];
3194 struct iface *iface = iface_lookup(br, cfg->name);
3195 const char *type = iface_get_type(cfg, br->cfg);
3200 } else if (!strcmp(type, "null")) {
3201 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3202 " may be removed in February 2013. Please email"
3203 " dev@openvswitch.org with concerns.",
3206 /* We will add new interfaces later. */
3212 /* Initializes 'oc' appropriately as a management service controller for
3215 * The caller must free oc->target when it is no longer needed. */
3217 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3218 struct ofproto_controller *oc)
3220 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3221 oc->max_backoff = 0;
3222 oc->probe_interval = 60;
3223 oc->band = OFPROTO_OUT_OF_BAND;
3225 oc->burst_limit = 0;
3226 oc->enable_async_msgs = true;
3230 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3232 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3233 struct ofproto_controller *oc)
3237 oc->target = c->target;
3238 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3239 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3240 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3241 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3242 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3243 oc->burst_limit = (c->controller_burst_limit
3244 ? *c->controller_burst_limit : 0);
3245 oc->enable_async_msgs = (!c->enable_async_messages
3246 || *c->enable_async_messages);
3247 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3248 if (dscp < 0 || dscp > 63) {
3249 dscp = DSCP_DEFAULT;
3254 /* Configures the IP stack for 'br''s local interface properly according to the
3255 * configuration in 'c'. */
3257 bridge_configure_local_iface_netdev(struct bridge *br,
3258 struct ovsrec_controller *c)
3260 struct netdev *netdev;
3261 struct in_addr mask, gateway;
3263 struct iface *local_iface;
3266 /* If there's no local interface or no IP address, give up. */
3267 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3268 if (!local_iface || !c->local_ip
3269 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3273 /* Bring up the local interface. */
3274 netdev = local_iface->netdev;
3275 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3277 /* Configure the IP address and netmask. */
3278 if (!c->local_netmask
3279 || !inet_pton(AF_INET, c->local_netmask, &mask)
3281 mask.s_addr = guess_netmask(ip.s_addr);
3283 if (!netdev_set_in4(netdev, ip, mask)) {
3284 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3285 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3288 /* Configure the default gateway. */
3289 if (c->local_gateway
3290 && inet_pton(AF_INET, c->local_gateway, &gateway)
3291 && gateway.s_addr) {
3292 if (!netdev_add_router(netdev, gateway)) {
3293 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3294 br->name, IP_ARGS(gateway.s_addr));
3299 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3300 * in either string are treated as equal to any number of slashes in the other,
3301 * e.g. "x///y" is equal to "x/y".
3303 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3304 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3305 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3306 * 'b' against a prefix of 'a'.
3309 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3311 const char *b_start = b;
3313 if (b - b_start >= b_stoplen) {
3315 } else if (*a != *b) {
3317 } else if (*a == '/') {
3318 a += strspn(a, "/");
3319 b += strspn(b, "/");
3320 } else if (*a == '\0') {
3330 bridge_configure_remotes(struct bridge *br,
3331 const struct sockaddr_in *managers, size_t n_managers)
3333 bool disable_in_band;
3335 struct ovsrec_controller **controllers;
3336 size_t n_controllers;
3338 enum ofproto_fail_mode fail_mode;
3340 struct ofproto_controller *ocs;
3344 /* Check if we should disable in-band control on this bridge. */
3345 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3348 /* Set OpenFlow queue ID for in-band control. */
3349 ofproto_set_in_band_queue(br->ofproto,
3350 smap_get_int(&br->cfg->other_config,
3351 "in-band-queue", -1));
3353 if (disable_in_band) {
3354 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3356 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3359 n_controllers = bridge_get_controllers(br, &controllers);
3361 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3364 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3365 for (i = 0; i < n_controllers; i++) {
3366 struct ovsrec_controller *c = controllers[i];
3368 if (!strncmp(c->target, "punix:", 6)
3369 || !strncmp(c->target, "unix:", 5)) {
3370 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3373 if (!strncmp(c->target, "unix:", 5)) {
3374 /* Connect to a listening socket */
3375 whitelist = xasprintf("unix:%s/", ovs_rundir());
3376 if (strchr(c->target, '/') &&
3377 !equal_pathnames(c->target, whitelist,
3378 strlen(whitelist))) {
3379 /* Absolute path specified, but not in ovs_rundir */
3380 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3381 "controller \"%s\" due to possibility for "
3382 "remote exploit. Instead, specify socket "
3383 "in whitelisted \"%s\" or connect to "
3384 "\"unix:%s/%s.mgmt\" (which is always "
3385 "available without special configuration).",
3386 br->name, c->target, whitelist,
3387 ovs_rundir(), br->name);
3392 whitelist = xasprintf("punix:%s/%s.controller",
3393 ovs_rundir(), br->name);
3394 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3395 /* Prevent remote ovsdb-server users from accessing
3396 * arbitrary Unix domain sockets and overwriting arbitrary
3398 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3399 "controller \"%s\" due to possibility of "
3400 "overwriting local files. Instead, specify "
3401 "whitelisted \"%s\" or connect to "
3402 "\"unix:%s/%s.mgmt\" (which is always "
3403 "available without special configuration).",
3404 br->name, c->target, whitelist,
3405 ovs_rundir(), br->name);
3414 bridge_configure_local_iface_netdev(br, c);
3415 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3416 if (disable_in_band) {
3417 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3422 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3423 bridge_get_allowed_versions(br));
3424 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3427 /* Set the fail-mode. */
3428 fail_mode = !br->cfg->fail_mode
3429 || !strcmp(br->cfg->fail_mode, "standalone")
3430 ? OFPROTO_FAIL_STANDALONE
3431 : OFPROTO_FAIL_SECURE;
3432 ofproto_set_fail_mode(br->ofproto, fail_mode);
3434 /* Configure OpenFlow controller connection snooping. */
3435 if (!ofproto_has_snoops(br->ofproto)) {
3439 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3440 ovs_rundir(), br->name));
3441 ofproto_set_snoops(br->ofproto, &snoops);
3442 sset_destroy(&snoops);
3447 bridge_configure_tables(struct bridge *br)
3449 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3453 n_tables = ofproto_get_n_tables(br->ofproto);
3455 for (i = 0; i < n_tables; i++) {
3456 struct ofproto_table_settings s;
3457 bool use_default_prefixes = true;
3460 s.max_flows = UINT_MAX;
3463 s.n_prefix_fields = 0;
3464 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3466 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3467 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3470 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3471 s.max_flows = *cfg->flow_limit;
3473 if (cfg->overflow_policy
3474 && !strcmp(cfg->overflow_policy, "evict")) {
3476 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3477 for (k = 0; k < cfg->n_groups; k++) {
3478 const char *string = cfg->groups[k];
3481 msg = mf_parse_subfield__(&s.groups[k], &string);
3483 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3484 "'groups' (%s)", br->name, i, msg);
3486 } else if (*string) {
3487 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3488 "element '%s' contains trailing garbage",
3489 br->name, i, cfg->groups[k]);
3495 /* Prefix lookup fields. */
3496 s.n_prefix_fields = 0;
3497 for (k = 0; k < cfg->n_prefixes; k++) {
3498 const char *name = cfg->prefixes[k];
3499 const struct mf_field *mf;
3501 if (strcmp(name, "none") == 0) {
3502 use_default_prefixes = false;
3503 s.n_prefix_fields = 0;
3506 mf = mf_from_name(name);
3508 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3512 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3513 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3514 "%s", br->name, name);
3517 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3518 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3519 "field not used: %s", br->name, name);
3522 use_default_prefixes = false;
3523 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3526 if (use_default_prefixes) {
3527 /* Use default values. */
3528 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3529 memcpy(s.prefix_fields, default_prefix_fields,
3530 sizeof default_prefix_fields);
3533 struct ds ds = DS_EMPTY_INITIALIZER;
3534 for (k = 0; k < s.n_prefix_fields; k++) {
3536 ds_put_char(&ds, ',');
3538 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3540 if (s.n_prefix_fields == 0) {
3541 ds_put_cstr(&ds, "none");
3543 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3544 br->name, i, ds_cstr(&ds));
3548 ofproto_configure_table(br->ofproto, i, &s);
3552 for (; j < br->cfg->n_flow_tables; j++) {
3553 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3554 "%"PRId64" not supported by this datapath", br->name,
3555 br->cfg->key_flow_tables[j]);
3560 bridge_configure_dp_desc(struct bridge *br)
3562 ofproto_set_dp_desc(br->ofproto,
3563 smap_get(&br->cfg->other_config, "dp-desc"));
3566 /* Port functions. */
3568 static struct port *
3569 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3573 port = xzalloc(sizeof *port);
3575 port->name = xstrdup(cfg->name);
3577 list_init(&port->ifaces);
3579 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3583 /* Deletes interfaces from 'port' that are no longer configured for it. */
3585 port_del_ifaces(struct port *port)
3587 struct iface *iface, *next;
3588 struct sset new_ifaces;
3591 /* Collect list of new interfaces. */
3592 sset_init(&new_ifaces);
3593 for (i = 0; i < port->cfg->n_interfaces; i++) {
3594 const char *name = port->cfg->interfaces[i]->name;
3595 const char *type = port->cfg->interfaces[i]->type;
3596 if (strcmp(type, "null")) {
3597 sset_add(&new_ifaces, name);
3601 /* Get rid of deleted interfaces. */
3602 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3603 if (!sset_contains(&new_ifaces, iface->name)) {
3604 iface_destroy(iface);
3608 sset_destroy(&new_ifaces);
3612 port_destroy(struct port *port)
3615 struct bridge *br = port->bridge;
3616 struct iface *iface, *next;
3619 ofproto_bundle_unregister(br->ofproto, port);
3622 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3623 iface_destroy__(iface);
3626 hmap_remove(&br->ports, &port->hmap_node);
3632 static struct port *
3633 port_lookup(const struct bridge *br, const char *name)
3637 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3639 if (!strcmp(port->name, name)) {
3647 enable_lacp(struct port *port, bool *activep)
3649 if (!port->cfg->lacp) {
3650 /* XXX when LACP implementation has been sufficiently tested, enable by
3651 * default and make active on bonded ports. */
3653 } else if (!strcmp(port->cfg->lacp, "off")) {
3655 } else if (!strcmp(port->cfg->lacp, "active")) {
3658 } else if (!strcmp(port->cfg->lacp, "passive")) {
3662 VLOG_WARN("port %s: unknown LACP mode %s",
3663 port->name, port->cfg->lacp);
3668 static struct lacp_settings *
3669 port_configure_lacp(struct port *port, struct lacp_settings *s)
3671 const char *lacp_time, *system_id;
3674 if (!enable_lacp(port, &s->active)) {
3678 s->name = port->name;
3680 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3682 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3683 ETH_ADDR_SCAN_ARGS(s->id))) {
3684 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3685 " address.", port->name, system_id);
3689 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3692 if (eth_addr_is_zero(s->id)) {
3693 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3697 /* Prefer bondable links if unspecified. */
3698 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3700 s->priority = (priority > 0 && priority <= UINT16_MAX
3702 : UINT16_MAX - !list_is_short(&port->ifaces));
3704 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3705 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3707 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3708 "lacp-fallback-ab", false);
3714 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3716 int priority, portid, key;
3718 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3719 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3721 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3723 if (portid <= 0 || portid > UINT16_MAX) {
3724 portid = ofp_to_u16(iface->ofp_port);
3727 if (priority <= 0 || priority > UINT16_MAX) {
3728 priority = UINT16_MAX;
3731 if (key < 0 || key > UINT16_MAX) {
3735 s->name = iface->name;
3737 s->priority = priority;
3742 port_configure_bond(struct port *port, struct bond_settings *s)
3744 const char *detect_s;
3745 struct iface *iface;
3746 int miimon_interval;
3748 s->name = port->name;
3750 if (port->cfg->bond_mode) {
3751 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3752 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3753 port->name, port->cfg->bond_mode,
3754 bond_mode_to_string(s->balance));
3757 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3759 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3760 * active-backup. At some point we should remove this warning. */
3761 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3762 " in previous versions, the default bond_mode was"
3763 " balance-slb", port->name,
3764 bond_mode_to_string(s->balance));
3766 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3767 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3768 "please use another bond type or disable flood_vlans",
3772 miimon_interval = smap_get_int(&port->cfg->other_config,
3773 "bond-miimon-interval", 0);
3774 if (miimon_interval <= 0) {
3775 miimon_interval = 200;
3778 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3779 if (!detect_s || !strcmp(detect_s, "carrier")) {
3780 miimon_interval = 0;
3781 } else if (strcmp(detect_s, "miimon")) {
3782 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3783 "defaulting to carrier", port->name, detect_s);
3784 miimon_interval = 0;
3787 s->up_delay = MAX(0, port->cfg->bond_updelay);
3788 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3789 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3790 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3791 "bond-rebalance-interval", 10000);
3792 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3793 s->rebalance_interval = 1000;
3796 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3797 "lacp-fallback-ab", false);
3799 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3800 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3804 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3805 * instead of obtaining it from the database. */
3807 port_is_synthetic(const struct port *port)
3809 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3812 /* Interface functions. */
3815 iface_is_internal(const struct ovsrec_interface *iface,
3816 const struct ovsrec_bridge *br)
3818 /* The local port and "internal" ports are always "internal". */
3819 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3822 /* Returns the correct network device type for interface 'iface' in bridge
3825 iface_get_type(const struct ovsrec_interface *iface,
3826 const struct ovsrec_bridge *br)
3830 /* The local port always has type "internal". Other ports take
3831 * their type from the database and default to "system" if none is
3833 if (iface_is_internal(iface, br)) {
3836 type = iface->type[0] ? iface->type : "system";
3839 return ofproto_port_open_type(br->datapath_type, type);
3843 iface_destroy__(struct iface *iface)
3846 struct port *port = iface->port;
3847 struct bridge *br = port->bridge;
3849 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3850 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3853 if (iface->ofp_port != OFPP_NONE) {
3854 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3857 list_remove(&iface->port_elem);
3858 hmap_remove(&br->iface_by_name, &iface->name_node);
3860 /* The user is changing configuration here, so netdev_remove needs to be
3861 * used as opposed to netdev_close */
3862 netdev_remove(iface->netdev);
3870 iface_destroy(struct iface *iface)
3873 struct port *port = iface->port;
3875 iface_destroy__(iface);
3876 if (list_is_empty(&port->ifaces)) {
3882 static struct iface *
3883 iface_lookup(const struct bridge *br, const char *name)
3885 struct iface *iface;
3887 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3888 &br->iface_by_name) {
3889 if (!strcmp(iface->name, name)) {
3897 static struct iface *
3898 iface_find(const char *name)
3900 const struct bridge *br;
3902 HMAP_FOR_EACH (br, node, &all_bridges) {
3903 struct iface *iface = iface_lookup(br, name);
3912 static struct iface *
3913 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3915 struct iface *iface;
3917 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3919 if (iface->ofp_port == ofp_port) {
3926 /* Set Ethernet address of 'iface', if one is specified in the configuration
3929 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3931 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3932 struct iface *hw_addr_iface;
3934 if (strcmp(iface->type, "internal")) {
3938 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3940 } else if (port->cfg->fake_bridge) {
3941 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3942 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3947 if (iface->ofp_port == OFPP_LOCAL) {
3948 VLOG_ERR("interface %s: ignoring mac in Interface record "
3949 "(use Bridge record to set local port's mac)",
3951 } else if (eth_addr_is_multicast(mac)) {
3952 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3955 int error = netdev_set_etheraddr(iface->netdev, mac);
3957 VLOG_ERR("interface %s: setting MAC failed (%s)",
3958 iface->name, ovs_strerror(error));
3964 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3966 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3968 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3969 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3970 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3974 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3975 * sets the "ofport" field to -1.
3977 * This is appropriate when 'if_cfg''s interface cannot be created or is
3978 * otherwise invalid. */
3980 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
3982 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3983 iface_set_ofport(if_cfg, OFPP_NONE);
3984 ovsrec_interface_set_error(if_cfg, errp);
3985 ovsrec_interface_set_status(if_cfg, NULL);
3986 ovsrec_interface_set_admin_state(if_cfg, NULL);
3987 ovsrec_interface_set_duplex(if_cfg, NULL);
3988 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3989 ovsrec_interface_set_link_state(if_cfg, NULL);
3990 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3991 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3992 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3993 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3994 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3995 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3996 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3997 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4002 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4004 union ovsdb_atom atom;
4006 atom.integer = target;
4007 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4011 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4013 struct ofpbuf queues_buf;
4015 ofpbuf_init(&queues_buf, 0);
4017 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
4018 netdev_set_qos(iface->netdev, NULL, NULL);
4020 const struct ovsdb_datum *queues;
4021 struct netdev_queue_dump dump;
4022 unsigned int queue_id;
4023 struct smap details;
4027 /* Configure top-level Qos for 'iface'. */
4028 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4030 /* Deconfigure queues that were deleted. */
4031 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4033 smap_init(&details);
4034 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4035 if (!queue_ids_include(queues, queue_id)) {
4036 netdev_delete_queue(iface->netdev, queue_id);
4039 smap_destroy(&details);
4041 /* Configure queues for 'iface'. */
4043 for (i = 0; i < qos->n_queues; i++) {
4044 const struct ovsrec_queue *queue = qos->value_queues[i];
4045 unsigned int queue_id = qos->key_queues[i];
4047 if (queue_id == 0) {
4051 if (queue->n_dscp == 1) {
4052 struct ofproto_port_queue *port_queue;
4054 port_queue = ofpbuf_put_uninit(&queues_buf,
4055 sizeof *port_queue);
4056 port_queue->queue = queue_id;
4057 port_queue->dscp = queue->dscp[0];
4060 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4063 struct smap details;
4065 smap_init(&details);
4066 netdev_set_queue(iface->netdev, 0, &details);
4067 smap_destroy(&details);
4071 if (iface->ofp_port != OFPP_NONE) {
4072 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
4073 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
4075 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4076 port_queues, n_queues);
4079 netdev_set_policing(iface->netdev,
4080 iface->cfg->ingress_policing_rate,
4081 iface->cfg->ingress_policing_burst);
4083 ofpbuf_uninit(&queues_buf);
4087 iface_configure_cfm(struct iface *iface)
4089 const struct ovsrec_interface *cfg = iface->cfg;
4090 const char *opstate_str;
4091 const char *cfm_ccm_vlan;
4092 struct cfm_settings s;
4093 struct smap netdev_args;
4095 if (!cfg->n_cfm_mpid) {
4096 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4100 s.check_tnl_key = false;
4101 smap_init(&netdev_args);
4102 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4103 const char *key = smap_get(&netdev_args, "key");
4104 const char *in_key = smap_get(&netdev_args, "in_key");
4106 s.check_tnl_key = (key && !strcmp(key, "flow"))
4107 || (in_key && !strcmp(in_key, "flow"));
4109 smap_destroy(&netdev_args);
4111 s.mpid = *cfg->cfm_mpid;
4112 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4113 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4114 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4116 if (s.interval <= 0) {
4120 if (!cfm_ccm_vlan) {
4122 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4123 s.ccm_vlan = CFM_RANDOM_VLAN;
4125 s.ccm_vlan = atoi(cfm_ccm_vlan);
4126 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4131 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4133 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4135 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4136 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4138 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4141 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4142 * instead of obtaining it from the database. */
4144 iface_is_synthetic(const struct iface *iface)
4146 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4150 iface_validate_ofport__(size_t n, int64_t *ofport)
4152 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4153 ? u16_to_ofp(*ofport)
4158 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4160 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4164 iface_pick_ofport(const struct ovsrec_interface *cfg)
4166 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4167 return (requested_ofport != OFPP_NONE
4169 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4172 /* Port mirroring. */
4174 static struct mirror *
4175 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4179 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4180 if (uuid_equals(uuid, &m->uuid)) {
4188 bridge_configure_mirrors(struct bridge *br)
4190 const struct ovsdb_datum *mc;
4191 unsigned long *flood_vlans;
4192 struct mirror *m, *next;
4195 /* Get rid of deleted mirrors. */
4196 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4197 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4198 union ovsdb_atom atom;
4200 atom.uuid = m->uuid;
4201 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4206 /* Add new mirrors and reconfigure existing ones. */
4207 for (i = 0; i < br->cfg->n_mirrors; i++) {
4208 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4209 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4211 m = mirror_create(br, cfg);
4214 if (!mirror_configure(m)) {
4219 /* Update flooded vlans (for RSPAN). */
4220 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4221 br->cfg->n_flood_vlans);
4222 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4223 bitmap_free(flood_vlans);
4226 static struct mirror *
4227 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4231 m = xzalloc(sizeof *m);
4232 m->uuid = cfg->header_.uuid;
4233 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4235 m->name = xstrdup(cfg->name);
4241 mirror_destroy(struct mirror *m)
4244 struct bridge *br = m->bridge;
4247 ofproto_mirror_unregister(br->ofproto, m);
4250 hmap_remove(&br->mirrors, &m->hmap_node);
4257 mirror_collect_ports(struct mirror *m,
4258 struct ovsrec_port **in_ports, int n_in_ports,
4259 void ***out_portsp, size_t *n_out_portsp)
4261 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4262 size_t n_out_ports = 0;
4265 for (i = 0; i < n_in_ports; i++) {
4266 const char *name = in_ports[i]->name;
4267 struct port *port = port_lookup(m->bridge, name);
4269 out_ports[n_out_ports++] = port;
4271 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4272 "port %s", m->bridge->name, m->name, name);
4275 *out_portsp = out_ports;
4276 *n_out_portsp = n_out_ports;
4280 mirror_configure(struct mirror *m)
4282 const struct ovsrec_mirror *cfg = m->cfg;
4283 struct ofproto_mirror_settings s;
4286 if (strcmp(cfg->name, m->name)) {
4288 m->name = xstrdup(cfg->name);
4292 /* Get output port or VLAN. */
4293 if (cfg->output_port) {
4294 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4295 if (!s.out_bundle) {
4296 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4297 m->bridge->name, m->name);
4300 s.out_vlan = UINT16_MAX;
4302 if (cfg->output_vlan) {
4303 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4304 "output vlan; ignoring output vlan",
4305 m->bridge->name, m->name);
4307 } else if (cfg->output_vlan) {
4308 /* The database should prevent invalid VLAN values. */
4309 s.out_bundle = NULL;
4310 s.out_vlan = *cfg->output_vlan;
4312 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4313 m->bridge->name, m->name);
4317 /* Get port selection. */
4318 if (cfg->select_all) {
4319 size_t n_ports = hmap_count(&m->bridge->ports);
4320 void **ports = xmalloc(n_ports * sizeof *ports);
4325 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4335 /* Get ports, dropping ports that don't exist.
4336 * The IDL ensures that there are no duplicates. */
4337 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4338 &s.srcs, &s.n_srcs);
4339 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4340 &s.dsts, &s.n_dsts);
4343 /* Get VLAN selection. */
4344 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4347 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4350 if (s.srcs != s.dsts) {
4359 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4361 * This is deprecated. It is only for compatibility with broken device drivers
4362 * in old versions of Linux that do not properly support VLANs when VLAN
4363 * devices are not used. When broken device drivers are no longer in
4364 * widespread use, we will delete these interfaces. */
4366 static struct ovsrec_port **recs;
4367 static size_t n_recs, allocated_recs;
4369 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4370 * splinters are reconfigured. */
4372 register_rec(struct ovsrec_port *rec)
4374 if (n_recs >= allocated_recs) {
4375 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4377 recs[n_recs++] = rec;
4380 /* Frees all of the ports registered with register_reg(). */
4382 free_registered_recs(void)
4386 for (i = 0; i < n_recs; i++) {
4387 struct ovsrec_port *port = recs[i];
4390 for (j = 0; j < port->n_interfaces; j++) {
4391 struct ovsrec_interface *iface = port->interfaces[j];
4396 smap_destroy(&port->other_config);
4397 free(port->interfaces);
4405 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4408 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4410 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4414 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4417 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4418 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4419 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4422 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4423 * use, returns NULL.
4425 * Updates 'vlan_splinters_enabled_anywhere'. */
4426 static unsigned long int *
4427 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4429 unsigned long int *splinter_vlans;
4430 struct sset splinter_ifaces;
4431 const char *real_dev_name;
4432 struct shash *real_devs;
4433 struct shash_node *node;
4437 /* Free space allocated for synthesized ports and interfaces, since we're
4438 * in the process of reconstructing all of them. */
4439 free_registered_recs();
4441 splinter_vlans = bitmap_allocate(4096);
4442 sset_init(&splinter_ifaces);
4443 vlan_splinters_enabled_anywhere = false;
4444 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4445 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4448 for (j = 0; j < br_cfg->n_ports; j++) {
4449 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4452 for (k = 0; k < port_cfg->n_interfaces; k++) {
4453 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4455 if (vlan_splinters_is_enabled(iface_cfg)) {
4456 vlan_splinters_enabled_anywhere = true;
4457 sset_add(&splinter_ifaces, iface_cfg->name);
4458 vlan_bitmap_from_array__(port_cfg->trunks,
4464 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4465 bitmap_set1(splinter_vlans, *port_cfg->tag);
4470 if (!vlan_splinters_enabled_anywhere) {
4471 free(splinter_vlans);
4472 sset_destroy(&splinter_ifaces);
4476 HMAP_FOR_EACH (br, node, &all_bridges) {
4478 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4482 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4483 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4484 * device to be created for it. */
4485 bitmap_set0(splinter_vlans, 0);
4486 bitmap_set0(splinter_vlans, 4095);
4488 /* Delete all VLAN devices that we don't need. */
4490 real_devs = vlandev_get_real_devs();
4491 SHASH_FOR_EACH (node, real_devs) {
4492 const struct vlan_real_dev *real_dev = node->data;
4493 const struct vlan_dev *vlan_dev;
4494 bool real_dev_has_splinters;
4496 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4498 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4499 if (!real_dev_has_splinters
4500 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4501 struct netdev *netdev;
4503 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4504 if (!netdev_get_in4(netdev, NULL, NULL) ||
4505 !netdev_get_in6(netdev, NULL)) {
4506 /* It has an IP address configured, so we don't own
4507 * it. Don't delete it. */
4509 vlandev_del(vlan_dev->name);
4511 netdev_close(netdev);
4518 /* Add all VLAN devices that we need. */
4519 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4522 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4523 if (!vlandev_get_name(real_dev_name, vid)) {
4524 vlandev_add(real_dev_name, vid);
4531 sset_destroy(&splinter_ifaces);
4533 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4534 free(splinter_vlans);
4537 return splinter_vlans;
4540 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4543 configure_splinter_port(struct port *port)
4545 struct ofproto *ofproto = port->bridge->ofproto;
4546 ofp_port_t realdev_ofp_port;
4547 const char *realdev_name;
4548 struct iface *vlandev, *realdev;
4550 ofproto_bundle_unregister(port->bridge->ofproto, port);
4552 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4555 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4556 realdev = iface_lookup(port->bridge, realdev_name);
4557 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4559 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4563 static struct ovsrec_port *
4564 synthesize_splinter_port(const char *real_dev_name,
4565 const char *vlan_dev_name, int vid)
4567 struct ovsrec_interface *iface;
4568 struct ovsrec_port *port;
4570 iface = xmalloc(sizeof *iface);
4571 ovsrec_interface_init(iface);
4572 iface->name = xstrdup(vlan_dev_name);
4573 iface->type = "system";
4575 port = xmalloc(sizeof *port);
4576 ovsrec_port_init(port);
4577 port->interfaces = xmemdup(&iface, sizeof iface);
4578 port->n_interfaces = 1;
4579 port->name = xstrdup(vlan_dev_name);
4580 port->vlan_mode = "splinter";
4581 port->tag = xmalloc(sizeof *port->tag);
4584 smap_add(&port->other_config, "realdev", real_dev_name);
4590 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4591 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4592 * 1-bit in the 'splinter_vlans' bitmap. */
4594 add_vlan_splinter_ports(struct bridge *br,
4595 const unsigned long int *splinter_vlans,
4596 struct shash *ports)
4600 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4601 * we're modifying 'ports'. */
4602 for (i = 0; i < br->cfg->n_ports; i++) {
4603 const char *name = br->cfg->ports[i]->name;
4604 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4607 for (j = 0; j < port_cfg->n_interfaces; j++) {
4608 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4610 if (vlan_splinters_is_enabled(iface_cfg)) {
4611 const char *real_dev_name;
4614 real_dev_name = iface_cfg->name;
4615 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4616 const char *vlan_dev_name;
4618 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4620 && !shash_find(ports, vlan_dev_name)) {
4621 shash_add(ports, vlan_dev_name,
4622 synthesize_splinter_port(
4623 real_dev_name, vlan_dev_name, vid));
4632 mirror_refresh_stats(struct mirror *m)
4634 struct ofproto *ofproto = m->bridge->ofproto;
4635 uint64_t tx_packets, tx_bytes;
4638 size_t stat_cnt = 0;
4640 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4641 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4645 if (tx_packets != UINT64_MAX) {
4646 keys[stat_cnt] = "tx_packets";
4647 values[stat_cnt] = tx_packets;
4650 if (tx_bytes != UINT64_MAX) {
4651 keys[stat_cnt] = "tx_bytes";
4652 values[stat_cnt] = tx_bytes;
4656 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);