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 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
196 * expensive. This can cause bridge_reconfigure() to take a long time during
197 * which no other work can be done. To deal with this problem, we limit port
198 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
199 * call to bridge_reconfigure(). If there is more work to do after the limit
200 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
201 * This allows the rest of the code to catch up on important things like
202 * forwarding packets. */
203 #define OFP_PORT_ACTION_WINDOW 10
205 static void add_del_bridges(const struct ovsrec_open_vswitch *);
206 static void bridge_run__(void);
207 static void bridge_create(const struct ovsrec_bridge *);
208 static void bridge_destroy(struct bridge *);
209 static struct bridge *bridge_lookup(const char *name);
210 static unixctl_cb_func bridge_unixctl_dump_flows;
211 static unixctl_cb_func bridge_unixctl_reconnect;
212 static size_t bridge_get_controllers(const struct bridge *br,
213 struct ovsrec_controller ***controllersp);
214 static void bridge_collect_wanted_ports(struct bridge *,
215 const unsigned long *splinter_vlans,
216 struct shash *wanted_ports);
217 static void bridge_delete_ofprotos(void);
218 static void bridge_delete_or_reconfigure_ports(struct bridge *);
219 static void bridge_del_ports(struct bridge *,
220 const struct shash *wanted_ports);
221 static void bridge_add_ports(struct bridge *,
222 const struct shash *wanted_ports);
224 static void bridge_configure_datapath_id(struct bridge *);
225 static void bridge_configure_netflow(struct bridge *);
226 static void bridge_configure_forward_bpdu(struct bridge *);
227 static void bridge_configure_mac_table(struct bridge *);
228 static void bridge_configure_mcast_snooping(struct bridge *);
229 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
230 static void bridge_configure_ipfix(struct bridge *);
231 static void bridge_configure_stp(struct bridge *);
232 static void bridge_configure_rstp(struct bridge *);
233 static void bridge_configure_tables(struct bridge *);
234 static void bridge_configure_dp_desc(struct bridge *);
235 static void bridge_configure_remotes(struct bridge *,
236 const struct sockaddr_in *managers,
238 static void bridge_pick_local_hw_addr(struct bridge *,
239 uint8_t ea[ETH_ADDR_LEN],
240 struct iface **hw_addr_iface);
241 static uint64_t bridge_pick_datapath_id(struct bridge *,
242 const uint8_t bridge_ea[ETH_ADDR_LEN],
243 struct iface *hw_addr_iface);
244 static uint64_t dpid_from_hash(const void *, size_t nbytes);
245 static bool bridge_has_bond_fake_iface(const struct bridge *,
247 static bool port_is_bond_fake_iface(const struct port *);
249 static unixctl_cb_func qos_unixctl_show;
251 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
252 static void port_del_ifaces(struct port *);
253 static void port_destroy(struct port *);
254 static struct port *port_lookup(const struct bridge *, const char *name);
255 static void port_configure(struct port *);
256 static struct lacp_settings *port_configure_lacp(struct port *,
257 struct lacp_settings *);
258 static void port_configure_bond(struct port *, struct bond_settings *);
259 static bool port_is_synthetic(const struct port *);
261 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
262 static void run_system_stats(void);
264 static void bridge_configure_mirrors(struct bridge *);
265 static struct mirror *mirror_create(struct bridge *,
266 const struct ovsrec_mirror *);
267 static void mirror_destroy(struct mirror *);
268 static bool mirror_configure(struct mirror *);
269 static void mirror_refresh_stats(struct mirror *);
271 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
272 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
273 const struct ovsrec_port *);
274 static bool iface_is_internal(const struct ovsrec_interface *iface,
275 const struct ovsrec_bridge *br);
276 static const char *iface_get_type(const struct ovsrec_interface *,
277 const struct ovsrec_bridge *);
278 static void iface_destroy(struct iface *);
279 static void iface_destroy__(struct iface *);
280 static struct iface *iface_lookup(const struct bridge *, const char *name);
281 static struct iface *iface_find(const char *name);
282 static struct iface *iface_from_ofp_port(const struct bridge *,
283 ofp_port_t ofp_port);
284 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
285 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
286 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
287 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
288 static void iface_configure_cfm(struct iface *);
289 static void iface_refresh_cfm_stats(struct iface *);
290 static void iface_refresh_stats(struct iface *);
291 static void iface_refresh_netdev_status(struct iface *);
292 static void iface_refresh_ofproto_status(struct iface *);
293 static bool iface_is_synthetic(const struct iface *);
294 static ofp_port_t iface_get_requested_ofp_port(
295 const struct ovsrec_interface *);
296 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
298 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
300 * This is deprecated. It is only for compatibility with broken device drivers
301 * in old versions of Linux that do not properly support VLANs when VLAN
302 * devices are not used. When broken device drivers are no longer in
303 * widespread use, we will delete these interfaces. */
305 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
306 static bool vlan_splinters_enabled_anywhere;
308 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
309 static unsigned long int *collect_splinter_vlans(
310 const struct ovsrec_open_vswitch *);
311 static void configure_splinter_port(struct port *);
312 static void add_vlan_splinter_ports(struct bridge *,
313 const unsigned long int *splinter_vlans,
314 struct shash *ports);
317 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
319 struct shash iface_hints;
320 static bool initialized = false;
327 shash_init(&iface_hints);
330 for (i = 0; i < cfg->n_bridges; i++) {
331 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
334 for (j = 0; j < br_cfg->n_ports; j++) {
335 struct ovsrec_port *port_cfg = br_cfg->ports[j];
338 for (k = 0; k < port_cfg->n_interfaces; k++) {
339 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
340 struct iface_hint *iface_hint;
342 iface_hint = xmalloc(sizeof *iface_hint);
343 iface_hint->br_name = br_cfg->name;
344 iface_hint->br_type = br_cfg->datapath_type;
345 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
347 shash_add(&iface_hints, if_cfg->name, iface_hint);
353 ofproto_init(&iface_hints);
355 shash_destroy_free_data(&iface_hints);
359 /* Public functions. */
361 /* Initializes the bridge module, configuring it to obtain its configuration
362 * from an OVSDB server accessed over 'remote', which should be a string in a
363 * form acceptable to ovsdb_idl_create(). */
365 bridge_init(const char *remote)
367 /* Create connection to database. */
368 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
369 idl_seqno = ovsdb_idl_get_seqno(idl);
370 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
371 ovsdb_idl_verify_write_only(idl);
373 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
374 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
375 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
376 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
378 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
379 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
381 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
382 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
383 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
384 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
385 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
386 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
388 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
389 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
390 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
391 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
392 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
408 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
409 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
410 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
411 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
412 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
413 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
414 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
416 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
417 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
418 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
419 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
421 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
423 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
425 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
426 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
428 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
429 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
430 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
431 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
433 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
434 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
435 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
436 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
437 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
439 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
441 /* Register unixctl commands. */
442 unixctl_command_register("qos/show", "interface", 1, 1,
443 qos_unixctl_show, NULL);
444 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
445 bridge_unixctl_dump_flows, NULL);
446 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
447 bridge_unixctl_reconnect, NULL);
459 struct bridge *br, *next_br;
461 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
464 ovsdb_idl_destroy(idl);
467 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
468 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
469 * responsible for freeing '*managersp' (with free()).
471 * You may be asking yourself "why does ovs-vswitchd care?", because
472 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
473 * should not be and in fact is not directly involved in that. But
474 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
475 * it has to tell in-band control where the managers are to enable that.
476 * (Thus, only managers connected in-band are collected.)
479 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
480 struct sockaddr_in **managersp, size_t *n_managersp)
482 struct sockaddr_in *managers = NULL;
483 size_t n_managers = 0;
487 /* Collect all of the potential targets from the "targets" columns of the
488 * rows pointed to by "manager_options", excluding any that are
491 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
492 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
494 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
495 sset_find_and_delete(&targets, m->target);
497 sset_add(&targets, m->target);
501 /* Now extract the targets' IP addresses. */
502 if (!sset_is_empty(&targets)) {
505 managers = xmalloc(sset_count(&targets) * sizeof *managers);
506 SSET_FOR_EACH (target, &targets) {
508 struct sockaddr_storage ss;
509 struct sockaddr_in in;
512 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
514 && sa.ss.ss_family == AF_INET) {
515 managers[n_managers++] = sa.in;
519 sset_destroy(&targets);
521 *managersp = managers;
522 *n_managersp = n_managers;
526 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
528 unsigned long int *splinter_vlans;
529 struct sockaddr_in *managers;
530 struct bridge *br, *next;
531 int sflow_bridge_number;
534 COVERAGE_INC(bridge_reconfigure);
536 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
537 OFPROTO_FLOW_LIMIT_DEFAULT));
538 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
539 OFPROTO_MAX_IDLE_DEFAULT));
540 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
542 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
545 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
546 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
548 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
549 * to 'ovs_cfg', with only very minimal configuration otherwise.
551 * This is mostly an update to bridge data structures. Nothing is pushed
552 * down to ofproto or lower layers. */
553 add_del_bridges(ovs_cfg);
554 splinter_vlans = collect_splinter_vlans(ovs_cfg);
555 HMAP_FOR_EACH (br, node, &all_bridges) {
556 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
557 bridge_del_ports(br, &br->wanted_ports);
559 free(splinter_vlans);
561 /* Start pushing configuration changes down to the ofproto layer:
563 * - Delete ofprotos that are no longer configured.
565 * - Delete ports that are no longer configured.
567 * - Reconfigure existing ports to their desired configurations, or
568 * delete them if not possible.
570 * We have to do all the deletions before we can do any additions, because
571 * the ports to be added might require resources that will be freed up by
572 * deletions (they might especially overlap in name). */
573 bridge_delete_ofprotos();
574 HMAP_FOR_EACH (br, node, &all_bridges) {
576 bridge_delete_or_reconfigure_ports(br);
580 /* Finish pushing configuration changes to the ofproto layer:
582 * - Create ofprotos that are missing.
584 * - Add ports that are missing. */
585 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
589 error = ofproto_create(br->name, br->type, &br->ofproto);
591 VLOG_ERR("failed to create bridge %s: %s", br->name,
592 ovs_strerror(error));
593 shash_destroy(&br->wanted_ports);
598 HMAP_FOR_EACH (br, node, &all_bridges) {
599 bridge_add_ports(br, &br->wanted_ports);
600 shash_destroy(&br->wanted_ports);
603 reconfigure_system_stats(ovs_cfg);
605 /* Complete the configuration. */
606 sflow_bridge_number = 0;
607 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
608 HMAP_FOR_EACH (br, node, &all_bridges) {
611 /* We need the datapath ID early to allow LACP ports to use it as the
612 * default system ID. */
613 bridge_configure_datapath_id(br);
615 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
618 port_configure(port);
620 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
621 iface_set_ofport(iface->cfg, iface->ofp_port);
622 /* Clear eventual previous errors */
623 ovsrec_interface_set_error(iface->cfg, NULL);
624 iface_configure_cfm(iface);
625 iface_configure_qos(iface, port->cfg->qos);
626 iface_set_mac(br, port, iface);
627 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
631 bridge_configure_mirrors(br);
632 bridge_configure_forward_bpdu(br);
633 bridge_configure_mac_table(br);
634 bridge_configure_mcast_snooping(br);
635 bridge_configure_remotes(br, managers, n_managers);
636 bridge_configure_netflow(br);
637 bridge_configure_sflow(br, &sflow_bridge_number);
638 bridge_configure_ipfix(br);
639 bridge_configure_stp(br);
640 bridge_configure_rstp(br);
641 bridge_configure_tables(br);
642 bridge_configure_dp_desc(br);
646 /* The ofproto-dpif provider does some final reconfiguration in its
647 * ->type_run() function. We have to call it before notifying the database
648 * client that reconfiguration is complete, otherwise there is a very
649 * narrow race window in which e.g. ofproto/trace will not recognize the
650 * new configuration (sometimes this causes unit test failures). */
654 /* Delete ofprotos which aren't configured or have the wrong type. Create
655 * ofprotos which don't exist but need to. */
657 bridge_delete_ofprotos(void)
664 /* Delete ofprotos with no bridge or with the wrong type. */
667 ofproto_enumerate_types(&types);
668 SSET_FOR_EACH (type, &types) {
671 ofproto_enumerate_names(type, &names);
672 SSET_FOR_EACH (name, &names) {
673 br = bridge_lookup(name);
674 if (!br || strcmp(type, br->type)) {
675 ofproto_delete(name, type);
679 sset_destroy(&names);
680 sset_destroy(&types);
684 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
686 if (*n >= *allocated) {
687 ports = x2nrealloc(ports, allocated, sizeof *ports);
689 ports[(*n)++] = port;
694 bridge_delete_or_reconfigure_ports(struct bridge *br)
696 struct ofproto_port ofproto_port;
697 struct ofproto_port_dump dump;
699 struct sset ofproto_ports;
700 struct port *port, *port_next;
702 /* List of "ofp_port"s to delete. We make a list instead of deleting them
703 * right away because ofproto implementations aren't necessarily able to
704 * iterate through a changing list of ports in an entirely robust way. */
711 sset_init(&ofproto_ports);
713 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
714 * that are not configured in the database. (This commonly happens when
715 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
717 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
718 * that have the wrong OpenFlow port number (and arrange to add them back
719 * with the correct OpenFlow port number). */
720 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
721 ofp_port_t requested_ofp_port;
724 sset_add(&ofproto_ports, ofproto_port.name);
726 iface = iface_lookup(br, ofproto_port.name);
728 /* No such iface is configured, so we should delete this
731 * As a corner case exception, keep the port if it's a bond fake
733 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
734 && !strcmp(ofproto_port.type, "internal")) {
740 if (strcmp(ofproto_port.type, iface->type)
741 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
742 /* The interface is the wrong type or can't be configured.
747 /* If the requested OpenFlow port for 'iface' changed, and it's not
748 * already the correct port, then we might want to temporarily delete
749 * this interface, so we can add it back again with the new OpenFlow
751 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
752 if (iface->ofp_port != OFPP_LOCAL &&
753 requested_ofp_port != OFPP_NONE &&
754 requested_ofp_port != iface->ofp_port) {
755 ofp_port_t victim_request;
756 struct iface *victim;
758 /* Check for an existing OpenFlow port currently occupying
759 * 'iface''s requested port number. If there isn't one, then
760 * delete this port. Otherwise we need to consider further. */
761 victim = iface_from_ofp_port(br, requested_ofp_port);
766 /* 'victim' is a port currently using 'iface''s requested port
767 * number. Unless 'victim' specifically requested that port
768 * number, too, then we can delete both 'iface' and 'victim'
769 * temporarily. (We'll add both of them back again later with new
770 * OpenFlow port numbers.)
772 * If 'victim' did request port number 'requested_ofp_port', just
773 * like 'iface', then that's a configuration inconsistency that we
774 * can't resolve. We might as well let it keep its current port
776 victim_request = iface_get_requested_ofp_port(victim->cfg);
777 if (victim_request != requested_ofp_port) {
778 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
779 iface_destroy(victim);
788 iface_destroy(iface);
789 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
791 for (i = 0; i < n; i++) {
792 ofproto_port_del(br->ofproto, del[i]);
796 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
797 * that we didn't see when we iterated through the datapath, i.e. ports
798 * that disappeared underneath use. This is an unusual situation, but it
799 * can happen in some cases:
801 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
802 * idea but could happen).
804 * - The port represented a device that disappeared, e.g. a tuntap
805 * device destroyed via "tunctl -d", a physical Ethernet device
806 * whose module was just unloaded via "rmmod", or a virtual NIC for a
807 * VM whose VM was just terminated. */
808 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
809 struct iface *iface, *iface_next;
811 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
812 if (!sset_contains(&ofproto_ports, iface->name)) {
813 iface_destroy__(iface);
817 if (list_is_empty(&port->ifaces)) {
821 sset_destroy(&ofproto_ports);
825 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
826 bool with_requested_port)
828 struct shash_node *port_node;
830 SHASH_FOR_EACH (port_node, wanted_ports) {
831 const struct ovsrec_port *port_cfg = port_node->data;
834 for (i = 0; i < port_cfg->n_interfaces; i++) {
835 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
836 ofp_port_t requested_ofp_port;
838 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
839 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
840 struct iface *iface = iface_lookup(br, iface_cfg->name);
843 iface_create(br, iface_cfg, port_cfg);
851 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
853 /* First add interfaces that request a particular port number. */
854 bridge_add_ports__(br, wanted_ports, true);
856 /* Then add interfaces that want automatic port number assignment.
857 * We add these afterward to avoid accidentally taking a specifically
858 * requested port number. */
859 bridge_add_ports__(br, wanted_ports, false);
863 port_configure(struct port *port)
865 const struct ovsrec_port *cfg = port->cfg;
866 struct bond_settings bond_settings;
867 struct lacp_settings lacp_settings;
868 struct ofproto_bundle_settings s;
871 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
872 configure_splinter_port(port);
881 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
882 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
883 s.slaves[s.n_slaves++] = iface->ofp_port;
888 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
892 /* Get VLAN trunks. */
895 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
899 if (cfg->vlan_mode) {
900 if (!strcmp(cfg->vlan_mode, "access")) {
901 s.vlan_mode = PORT_VLAN_ACCESS;
902 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
903 s.vlan_mode = PORT_VLAN_TRUNK;
904 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
905 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
906 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
907 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
909 /* This "can't happen" because ovsdb-server should prevent it. */
910 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
911 "back to trunk mode", port->name, cfg->vlan_mode);
912 s.vlan_mode = PORT_VLAN_TRUNK;
916 s.vlan_mode = PORT_VLAN_ACCESS;
918 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
922 s.vlan_mode = PORT_VLAN_TRUNK;
925 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
928 /* Get LACP settings. */
929 s.lacp = port_configure_lacp(port, &lacp_settings);
933 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
934 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
935 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
938 s.lacp_slaves = NULL;
941 /* Get bond settings. */
942 if (s.n_slaves > 1) {
943 s.bond = &bond_settings;
944 port_configure_bond(port, &bond_settings);
947 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
948 netdev_set_miimon_interval(iface->netdev, 0);
953 ofproto_bundle_register(port->bridge->ofproto, port, &s);
961 /* Pick local port hardware address and datapath ID for 'br'. */
963 bridge_configure_datapath_id(struct bridge *br)
965 uint8_t ea[ETH_ADDR_LEN];
967 struct iface *local_iface;
968 struct iface *hw_addr_iface;
971 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
972 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
974 int error = netdev_set_etheraddr(local_iface->netdev, ea);
976 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
977 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
978 "Ethernet address: %s",
979 br->name, ovs_strerror(error));
982 memcpy(br->ea, ea, ETH_ADDR_LEN);
984 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
985 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
986 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
987 ofproto_set_datapath_id(br->ofproto, dpid);
990 dpid_string = xasprintf("%016"PRIx64, dpid);
991 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
995 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
998 bridge_get_allowed_versions(struct bridge *br)
1000 if (!br->cfg->n_protocols)
1003 return ofputil_versions_from_strings(br->cfg->protocols,
1004 br->cfg->n_protocols);
1007 /* Set NetFlow configuration on 'br'. */
1009 bridge_configure_netflow(struct bridge *br)
1011 struct ovsrec_netflow *cfg = br->cfg->netflow;
1012 struct netflow_options opts;
1015 ofproto_set_netflow(br->ofproto, NULL);
1019 memset(&opts, 0, sizeof opts);
1021 /* Get default NetFlow configuration from datapath.
1022 * Apply overrides from 'cfg'. */
1023 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1024 if (cfg->engine_type) {
1025 opts.engine_type = *cfg->engine_type;
1027 if (cfg->engine_id) {
1028 opts.engine_id = *cfg->engine_id;
1031 /* Configure active timeout interval. */
1032 opts.active_timeout = cfg->active_timeout;
1033 if (!opts.active_timeout) {
1034 opts.active_timeout = -1;
1035 } else if (opts.active_timeout < 0) {
1036 VLOG_WARN("bridge %s: active timeout interval set to negative "
1037 "value, using default instead (%d seconds)", br->name,
1038 NF_ACTIVE_TIMEOUT_DEFAULT);
1039 opts.active_timeout = -1;
1042 /* Add engine ID to interface number to disambiguate bridgs? */
1043 opts.add_id_to_iface = cfg->add_id_to_interface;
1044 if (opts.add_id_to_iface) {
1045 if (opts.engine_id > 0x7f) {
1046 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1047 "another vswitch, choose an engine id less than 128",
1050 if (hmap_count(&br->ports) > 508) {
1051 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1052 "another port when more than 508 ports are used",
1058 sset_init(&opts.collectors);
1059 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1062 if (ofproto_set_netflow(br->ofproto, &opts)) {
1063 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1065 sset_destroy(&opts.collectors);
1068 /* Set sFlow configuration on 'br'. */
1070 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1072 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1073 struct ovsrec_controller **controllers;
1074 struct ofproto_sflow_options oso;
1075 size_t n_controllers;
1079 ofproto_set_sflow(br->ofproto, NULL);
1083 memset(&oso, 0, sizeof oso);
1085 sset_init(&oso.targets);
1086 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1088 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1089 if (cfg->sampling) {
1090 oso.sampling_rate = *cfg->sampling;
1093 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1095 oso.polling_interval = *cfg->polling;
1098 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1100 oso.header_len = *cfg->header;
1103 oso.sub_id = (*sflow_bridge_number)++;
1104 oso.agent_device = cfg->agent;
1106 oso.control_ip = NULL;
1107 n_controllers = bridge_get_controllers(br, &controllers);
1108 for (i = 0; i < n_controllers; i++) {
1109 if (controllers[i]->local_ip) {
1110 oso.control_ip = controllers[i]->local_ip;
1114 ofproto_set_sflow(br->ofproto, &oso);
1116 sset_destroy(&oso.targets);
1119 /* Returns whether a IPFIX row is valid. */
1121 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1123 return ipfix && ipfix->n_targets > 0;
1126 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1128 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1129 const struct bridge *br)
1131 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1134 /* Set IPFIX configuration on 'br'. */
1136 bridge_configure_ipfix(struct bridge *br)
1138 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1139 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1140 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1141 struct ofproto_ipfix_bridge_exporter_options be_opts;
1142 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1143 size_t n_fe_opts = 0;
1145 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1146 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1151 if (!valid_be_cfg && n_fe_opts == 0) {
1152 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1157 memset(&be_opts, 0, sizeof be_opts);
1159 sset_init(&be_opts.targets);
1160 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1162 if (be_cfg->sampling) {
1163 be_opts.sampling_rate = *be_cfg->sampling;
1165 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1167 if (be_cfg->obs_domain_id) {
1168 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1170 if (be_cfg->obs_point_id) {
1171 be_opts.obs_point_id = *be_cfg->obs_point_id;
1173 if (be_cfg->cache_active_timeout) {
1174 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1176 if (be_cfg->cache_max_flows) {
1177 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1180 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1181 "enable-tunnel-sampling", true);
1183 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1184 "enable-input-sampling", false);
1186 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1187 "enable-output-sampling", false);
1190 if (n_fe_opts > 0) {
1191 struct ofproto_ipfix_flow_exporter_options *opts;
1192 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1194 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1195 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1196 opts->collector_set_id = fe_cfg->id;
1197 sset_init(&opts->targets);
1198 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1199 fe_cfg->ipfix->n_targets);
1200 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1201 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1202 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1203 ? *fe_cfg->ipfix->cache_max_flows : 0;
1209 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1213 sset_destroy(&be_opts.targets);
1216 if (n_fe_opts > 0) {
1217 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1219 for (i = 0; i < n_fe_opts; i++) {
1220 sset_destroy(&opts->targets);
1228 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1229 struct ofproto_port_stp_settings *port_s,
1230 int *port_num_counter, unsigned long *port_num_bitmap)
1232 const char *config_str;
1233 struct iface *iface;
1235 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1236 port_s->enable = false;
1239 port_s->enable = true;
1242 /* STP over bonds is not supported. */
1243 if (!list_is_singleton(&port->ifaces)) {
1244 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1246 port_s->enable = false;
1250 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1252 /* Internal ports shouldn't participate in spanning tree, so
1254 if (!strcmp(iface->type, "internal")) {
1255 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1256 port_s->enable = false;
1260 /* STP on mirror output ports is not supported. */
1261 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1262 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1263 port_s->enable = false;
1267 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1269 unsigned long int port_num = strtoul(config_str, NULL, 0);
1270 int port_idx = port_num - 1;
1272 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1273 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1274 port_s->enable = false;
1278 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1279 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1280 port->name, port_num);
1281 port_s->enable = false;
1284 bitmap_set1(port_num_bitmap, port_idx);
1285 port_s->port_num = port_idx;
1287 if (*port_num_counter >= STP_MAX_PORTS) {
1288 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1289 port_s->enable = false;
1293 port_s->port_num = (*port_num_counter)++;
1296 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1298 port_s->path_cost = strtoul(config_str, NULL, 10);
1300 enum netdev_features current;
1303 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1304 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1305 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1308 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1310 port_s->priority = strtoul(config_str, NULL, 0);
1312 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1317 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1318 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1320 const char *config_str;
1321 struct iface *iface;
1323 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1324 port_s->enable = false;
1327 port_s->enable = true;
1330 /* RSTP over bonds is not supported. */
1331 if (!list_is_singleton(&port->ifaces)) {
1332 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1334 port_s->enable = false;
1338 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1340 /* Internal ports shouldn't participate in spanning tree, so
1342 if (!strcmp(iface->type, "internal")) {
1343 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1344 port_s->enable = false;
1348 /* RSTP on mirror output ports is not supported. */
1349 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1350 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1351 port_s->enable = false;
1355 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1357 unsigned long int port_num = strtoul(config_str, NULL, 0);
1358 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1359 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1360 port_s->enable = false;
1363 port_s->port_num = port_num;
1365 if (*port_num_counter >= RSTP_MAX_PORTS) {
1366 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1367 port_s->enable = false;
1370 /* If rstp-port-num is not specified, use 0.
1371 * rstp_port_set_port_number() will look for the first free one. */
1372 port_s->port_num = 0;
1375 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1377 port_s->path_cost = strtoul(config_str, NULL, 10);
1379 enum netdev_features current;
1382 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1383 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1384 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1387 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1389 port_s->priority = strtoul(config_str, NULL, 0);
1391 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1394 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1395 "rstp-port-admin-edge", false);
1396 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1397 "rstp-port-auto-edge", true);
1398 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1399 "rstp-port-mcheck", false);
1402 /* Set spanning tree configuration on 'br'. */
1404 bridge_configure_stp(struct bridge *br)
1406 struct ofproto_rstp_status rstp_status;
1408 ofproto_get_rstp_status(br->ofproto, &rstp_status);
1409 if (!br->cfg->stp_enable) {
1410 ofproto_set_stp(br->ofproto, NULL);
1411 } else if (rstp_status.enabled) {
1412 /* Do not activate STP if RSTP is enabled. */
1413 VLOG_ERR("STP cannot be enabled if RSTP is running.");
1414 ofproto_set_stp(br->ofproto, NULL);
1415 ovsrec_bridge_set_stp_enable(br->cfg, false);
1417 struct ofproto_stp_settings br_s;
1418 const char *config_str;
1420 int port_num_counter;
1421 unsigned long *port_num_bitmap;
1423 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1425 uint8_t ea[ETH_ADDR_LEN];
1427 if (eth_addr_from_string(config_str, ea)) {
1428 br_s.system_id = eth_addr_to_uint64(ea);
1430 br_s.system_id = eth_addr_to_uint64(br->ea);
1431 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1432 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1435 br_s.system_id = eth_addr_to_uint64(br->ea);
1438 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1440 br_s.priority = strtoul(config_str, NULL, 0);
1442 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1445 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1447 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1449 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1452 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1454 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1456 br_s.max_age = STP_DEFAULT_MAX_AGE;
1459 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1461 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1463 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1466 /* Configure STP on the bridge. */
1467 if (ofproto_set_stp(br->ofproto, &br_s)) {
1468 VLOG_ERR("bridge %s: could not enable STP", br->name);
1472 /* Users must either set the port number with the "stp-port-num"
1473 * configuration on all ports or none. If manual configuration
1474 * is not done, then we allocate them sequentially. */
1475 port_num_counter = 0;
1476 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1477 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1478 struct ofproto_port_stp_settings port_s;
1479 struct iface *iface;
1481 port_configure_stp(br->ofproto, port, &port_s,
1482 &port_num_counter, port_num_bitmap);
1484 /* As bonds are not supported, just apply configuration to
1485 * all interfaces. */
1486 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1487 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1489 VLOG_ERR("port %s: could not enable STP", port->name);
1495 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1496 && port_num_counter) {
1497 VLOG_ERR("bridge %s: must manually configure all STP port "
1498 "IDs or none, disabling", br->name);
1499 ofproto_set_stp(br->ofproto, NULL);
1501 bitmap_free(port_num_bitmap);
1506 bridge_configure_rstp(struct bridge *br)
1508 struct ofproto_stp_status stp_status;
1510 ofproto_get_stp_status(br->ofproto, &stp_status);
1511 if (!br->cfg->rstp_enable) {
1512 ofproto_set_rstp(br->ofproto, NULL);
1513 } else if (stp_status.enabled) {
1514 /* Do not activate RSTP if STP is enabled. */
1515 VLOG_ERR("RSTP cannot be enabled if STP is running.");
1516 ofproto_set_rstp(br->ofproto, NULL);
1517 ovsrec_bridge_set_rstp_enable(br->cfg, false);
1519 struct ofproto_rstp_settings br_s;
1520 const char *config_str;
1522 int port_num_counter;
1524 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1526 uint8_t ea[ETH_ADDR_LEN];
1528 if (eth_addr_from_string(config_str, ea)) {
1529 br_s.address = eth_addr_to_uint64(ea);
1532 br_s.address = eth_addr_to_uint64(br->ea);
1533 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1534 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1538 br_s.address = eth_addr_to_uint64(br->ea);
1541 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1543 br_s.priority = strtoul(config_str, NULL, 0);
1545 br_s.priority = RSTP_DEFAULT_PRIORITY;
1548 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1550 br_s.ageing_time = strtoul(config_str, NULL, 0);
1552 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1555 config_str = smap_get(&br->cfg->other_config,
1556 "rstp-force-protocol-version");
1558 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1560 br_s.force_protocol_version = FPV_DEFAULT;
1563 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1565 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1567 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1570 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1572 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1574 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1577 config_str = smap_get(&br->cfg->other_config,
1578 "rstp-transmit-hold-count");
1580 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1582 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1585 /* Configure RSTP on the bridge. */
1586 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1587 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1591 port_num_counter = 0;
1592 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1593 struct ofproto_port_rstp_settings port_s;
1594 struct iface *iface;
1596 port_configure_rstp(br->ofproto, port, &port_s,
1599 /* As bonds are not supported, just apply configuration to
1600 * all interfaces. */
1601 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1602 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1604 VLOG_ERR("port %s: could not enable RSTP", port->name);
1613 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1615 const struct port *port = port_lookup(br, name);
1616 return port && port_is_bond_fake_iface(port);
1620 port_is_bond_fake_iface(const struct port *port)
1622 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1626 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1628 struct bridge *br, *next;
1629 struct shash new_br;
1632 /* Collect new bridges' names and types. */
1633 shash_init(&new_br);
1634 for (i = 0; i < cfg->n_bridges; i++) {
1635 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1636 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1638 if (strchr(br_cfg->name, '/')) {
1639 /* Prevent remote ovsdb-server users from accessing arbitrary
1640 * directories, e.g. consider a bridge named "../../../etc/". */
1641 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1643 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1644 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1648 /* Get rid of deleted bridges or those whose types have changed.
1649 * Update 'cfg' of bridges that still exist. */
1650 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1651 br->cfg = shash_find_data(&new_br, br->name);
1652 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1653 br->cfg->datapath_type))) {
1658 /* Add new bridges. */
1659 for (i = 0; i < cfg->n_bridges; i++) {
1660 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1661 struct bridge *br = bridge_lookup(br_cfg->name);
1663 bridge_create(br_cfg);
1667 shash_destroy(&new_br);
1670 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1671 * Returns 0 if successful, otherwise a positive errno value. */
1673 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1674 struct netdev *netdev, char **errp)
1676 return netdev_set_config(netdev, &iface_cfg->options, errp);
1679 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1680 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1682 * If successful, returns 0 and stores the network device in '*netdevp'. On
1683 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1685 iface_do_create(const struct bridge *br,
1686 const struct ovsrec_interface *iface_cfg,
1687 const struct ovsrec_port *port_cfg,
1688 ofp_port_t *ofp_portp, struct netdev **netdevp,
1691 struct netdev *netdev = NULL;
1694 if (netdev_is_reserved_name(iface_cfg->name)) {
1695 VLOG_WARN("could not create interface %s, name is reserved",
1701 error = netdev_open(iface_cfg->name,
1702 iface_get_type(iface_cfg, br->cfg), &netdev);
1704 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1705 iface_cfg->name, ovs_strerror(error));
1709 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1714 *ofp_portp = iface_pick_ofport(iface_cfg);
1715 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1720 VLOG_INFO("bridge %s: added interface %s on port %d",
1721 br->name, iface_cfg->name, *ofp_portp);
1723 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1724 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1732 netdev_close(netdev);
1736 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1737 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1738 * automatically allocated for the iface. Takes ownership of and
1739 * deallocates 'if_cfg'.
1741 * Return true if an iface is successfully created, false otherwise. */
1743 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1744 const struct ovsrec_port *port_cfg)
1746 struct netdev *netdev;
1747 struct iface *iface;
1748 ofp_port_t ofp_port;
1753 /* Do the bits that can fail up front. */
1754 ovs_assert(!iface_lookup(br, iface_cfg->name));
1755 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1757 iface_clear_db_record(iface_cfg, errp);
1762 /* Get or create the port structure. */
1763 port = port_lookup(br, port_cfg->name);
1765 port = port_create(br, port_cfg);
1768 /* Create the iface structure. */
1769 iface = xzalloc(sizeof *iface);
1770 list_push_back(&port->ifaces, &iface->port_elem);
1771 hmap_insert(&br->iface_by_name, &iface->name_node,
1772 hash_string(iface_cfg->name, 0));
1774 iface->name = xstrdup(iface_cfg->name);
1775 iface->ofp_port = ofp_port;
1776 iface->netdev = netdev;
1777 iface->type = iface_get_type(iface_cfg, br->cfg);
1778 iface->cfg = iface_cfg;
1779 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1780 hash_ofp_port(ofp_port));
1782 /* Populate initial status in database. */
1783 iface_refresh_stats(iface);
1784 iface_refresh_netdev_status(iface);
1786 /* Add bond fake iface if necessary. */
1787 if (port_is_bond_fake_iface(port)) {
1788 struct ofproto_port ofproto_port;
1790 if (ofproto_port_query_by_name(br->ofproto, port->name,
1792 struct netdev *netdev;
1795 error = netdev_open(port->name, "internal", &netdev);
1797 ofp_port_t fake_ofp_port = OFPP_NONE;
1798 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1799 netdev_close(netdev);
1801 VLOG_WARN("could not open network device %s (%s)",
1802 port->name, ovs_strerror(error));
1805 /* Already exists, nothing to do. */
1806 ofproto_port_destroy(&ofproto_port);
1813 /* Set forward BPDU option. */
1815 bridge_configure_forward_bpdu(struct bridge *br)
1817 ofproto_set_forward_bpdu(br->ofproto,
1818 smap_get_bool(&br->cfg->other_config,
1823 /* Set MAC learning table configuration for 'br'. */
1825 bridge_configure_mac_table(struct bridge *br)
1827 const char *idle_time_str;
1830 const char *mac_table_size_str;
1833 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1834 idle_time = (idle_time_str && atoi(idle_time_str)
1835 ? atoi(idle_time_str)
1836 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1838 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1839 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1840 ? atoi(mac_table_size_str)
1843 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1846 /* Set multicast snooping table configuration for 'br'. */
1848 bridge_configure_mcast_snooping(struct bridge *br)
1850 if (!br->cfg->mcast_snooping_enable) {
1851 ofproto_set_mcast_snooping(br->ofproto, NULL);
1854 struct ofproto_mcast_snooping_settings br_s;
1855 const char *idle_time_str;
1856 const char *max_entries_str;
1858 idle_time_str = smap_get(&br->cfg->other_config,
1859 "mcast-snooping-aging-time");
1860 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1861 ? atoi(idle_time_str)
1862 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1864 max_entries_str = smap_get(&br->cfg->other_config,
1865 "mcast-snooping-table-size");
1866 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1867 ? atoi(max_entries_str)
1868 : MCAST_DEFAULT_MAX_ENTRIES);
1870 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1871 "mcast-snooping-disable-flood-unregistered",
1874 /* Configure multicast snooping on the bridge */
1875 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1876 VLOG_ERR("bridge %s: could not enable multicast snooping",
1881 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1882 bool flood = smap_get_bool(&port->cfg->other_config,
1883 "mcast-snooping-flood", false);
1884 if (ofproto_port_set_mcast_snooping(br->ofproto, port, flood)) {
1885 VLOG_ERR("port %s: could not configure mcast snooping",
1893 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1894 const struct port *fake_br, struct iface **hw_addr_iface)
1896 struct hmapx mirror_output_ports;
1898 bool found_addr = false;
1902 /* Mirror output ports don't participate in picking the local hardware
1903 * address. ofproto can't help us find out whether a given port is a
1904 * mirror output because we haven't configured mirrors yet, so we need to
1905 * accumulate them ourselves. */
1906 hmapx_init(&mirror_output_ports);
1907 for (i = 0; i < br->cfg->n_mirrors; i++) {
1908 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1909 if (m->output_port) {
1910 hmapx_add(&mirror_output_ports, m->output_port);
1914 /* Otherwise choose the minimum non-local MAC address among all of the
1916 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1917 uint8_t iface_ea[ETH_ADDR_LEN];
1918 struct iface *candidate;
1919 struct iface *iface;
1921 /* Mirror output ports don't participate. */
1922 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1926 /* Choose the MAC address to represent the port. */
1928 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1929 /* Find the interface with this Ethernet address (if any) so that
1930 * we can provide the correct devname to the caller. */
1931 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1932 uint8_t candidate_ea[ETH_ADDR_LEN];
1933 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1934 && eth_addr_equals(iface_ea, candidate_ea)) {
1939 /* Choose the interface whose MAC address will represent the port.
1940 * The Linux kernel bonding code always chooses the MAC address of
1941 * the first slave added to a bond, and the Fedora networking
1942 * scripts always add slaves to a bond in alphabetical order, so
1943 * for compatibility we choose the interface with the name that is
1944 * first in alphabetical order. */
1945 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1946 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1951 /* The local port doesn't count (since we're trying to choose its
1952 * MAC address anyway). */
1953 if (iface->ofp_port == OFPP_LOCAL) {
1957 /* For fake bridges we only choose from ports with the same tag */
1958 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1959 if (!port->cfg->tag) {
1962 if (*port->cfg->tag != *fake_br->cfg->tag) {
1968 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1974 /* Compare against our current choice. */
1975 if (!eth_addr_is_multicast(iface_ea) &&
1976 !eth_addr_is_local(iface_ea) &&
1977 !eth_addr_is_reserved(iface_ea) &&
1978 !eth_addr_is_zero(iface_ea) &&
1979 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1981 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1982 *hw_addr_iface = iface;
1988 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1989 *hw_addr_iface = NULL;
1992 hmapx_destroy(&mirror_output_ports);
1996 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1997 struct iface **hw_addr_iface)
2000 *hw_addr_iface = NULL;
2002 /* Did the user request a particular MAC? */
2003 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2004 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2005 if (eth_addr_is_multicast(ea)) {
2006 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2007 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2008 } else if (eth_addr_is_zero(ea)) {
2009 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2015 /* Find a local hw address */
2016 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2019 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2020 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2021 * an interface on 'br', then that interface must be passed in as
2022 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2023 * 'hw_addr_iface' must be passed in as a null pointer. */
2025 bridge_pick_datapath_id(struct bridge *br,
2026 const uint8_t bridge_ea[ETH_ADDR_LEN],
2027 struct iface *hw_addr_iface)
2030 * The procedure for choosing a bridge MAC address will, in the most
2031 * ordinary case, also choose a unique MAC that we can use as a datapath
2032 * ID. In some special cases, though, multiple bridges will end up with
2033 * the same MAC address. This is OK for the bridges, but it will confuse
2034 * the OpenFlow controller, because each datapath needs a unique datapath
2037 * Datapath IDs must be unique. It is also very desirable that they be
2038 * stable from one run to the next, so that policy set on a datapath
2041 const char *datapath_id;
2044 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2045 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2049 if (!hw_addr_iface) {
2051 * A purely internal bridge, that is, one that has no non-virtual
2052 * network devices on it at all, is difficult because it has no
2053 * natural unique identifier at all.
2055 * When the host is a XenServer, we handle this case by hashing the
2056 * host's UUID with the name of the bridge. Names of bridges are
2057 * persistent across XenServer reboots, although they can be reused if
2058 * an internal network is destroyed and then a new one is later
2059 * created, so this is fairly effective.
2061 * When the host is not a XenServer, we punt by using a random MAC
2062 * address on each run.
2064 const char *host_uuid = xenserver_get_host_uuid();
2066 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2067 dpid = dpid_from_hash(combined, strlen(combined));
2073 return eth_addr_to_uint64(bridge_ea);
2077 dpid_from_hash(const void *data, size_t n)
2079 uint8_t hash[SHA1_DIGEST_SIZE];
2081 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2082 sha1_bytes(data, n, hash);
2083 eth_addr_mark_random(hash);
2084 return eth_addr_to_uint64(hash);
2088 iface_refresh_netdev_status(struct iface *iface)
2092 enum netdev_features current;
2093 enum netdev_flags flags;
2094 const char *link_state;
2095 uint8_t mac[ETH_ADDR_LEN];
2096 int64_t bps, mtu_64, ifindex64, link_resets;
2099 if (iface_is_synthetic(iface)) {
2103 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2104 && !status_txn_try_again) {
2108 iface->change_seq = netdev_get_change_seq(iface->netdev);
2112 if (!netdev_get_status(iface->netdev, &smap)) {
2113 ovsrec_interface_set_status(iface->cfg, &smap);
2115 ovsrec_interface_set_status(iface->cfg, NULL);
2118 smap_destroy(&smap);
2120 error = netdev_get_flags(iface->netdev, &flags);
2122 const char *state = flags & NETDEV_UP ? "up" : "down";
2124 ovsrec_interface_set_admin_state(iface->cfg, state);
2126 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2129 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2130 ovsrec_interface_set_link_state(iface->cfg, link_state);
2132 link_resets = netdev_get_carrier_resets(iface->netdev);
2133 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2135 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2136 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2138 ovsrec_interface_set_duplex(iface->cfg,
2139 netdev_features_is_full_duplex(current)
2141 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2143 ovsrec_interface_set_duplex(iface->cfg, NULL);
2144 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2147 error = netdev_get_mtu(iface->netdev, &mtu);
2150 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2152 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2155 error = netdev_get_etheraddr(iface->netdev, mac);
2157 char mac_string[32];
2159 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2160 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2162 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2165 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2166 * if there is no valid ifindex number. */
2167 ifindex64 = netdev_get_ifindex(iface->netdev);
2168 if (ifindex64 < 0) {
2171 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2175 iface_refresh_ofproto_status(struct iface *iface)
2179 if (iface_is_synthetic(iface)) {
2183 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2187 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2189 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2192 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2194 || status_txn_try_again) {
2195 iface_refresh_cfm_stats(iface);
2198 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2200 || status_txn_try_again) {
2204 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2205 iface->ofp_port, &smap);
2206 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2207 smap_destroy(&smap);
2211 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2214 iface_refresh_cfm_stats(struct iface *iface)
2216 const struct ovsrec_interface *cfg = iface->cfg;
2217 struct cfm_status status;
2220 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2221 iface->ofp_port, &status);
2223 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2224 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2225 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2226 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2227 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2228 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2230 const char *reasons[CFM_FAULT_N_REASONS];
2231 int64_t cfm_health = status.health;
2232 int64_t cfm_flap_count = status.flap_count;
2233 bool faulted = status.faults != 0;
2236 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2239 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2240 int reason = 1 << i;
2241 if (status.faults & reason) {
2242 reasons[j++] = cfm_fault_reason_to_str(reason);
2245 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
2247 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2249 if (status.remote_opstate >= 0) {
2250 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2251 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2253 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2256 ovsrec_interface_set_cfm_remote_mpids(cfg,
2257 (const int64_t *)status.rmps,
2259 if (cfm_health >= 0) {
2260 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2262 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2270 iface_refresh_stats(struct iface *iface)
2272 #define IFACE_STATS \
2273 IFACE_STAT(rx_packets, "rx_packets") \
2274 IFACE_STAT(tx_packets, "tx_packets") \
2275 IFACE_STAT(rx_bytes, "rx_bytes") \
2276 IFACE_STAT(tx_bytes, "tx_bytes") \
2277 IFACE_STAT(rx_dropped, "rx_dropped") \
2278 IFACE_STAT(tx_dropped, "tx_dropped") \
2279 IFACE_STAT(rx_errors, "rx_errors") \
2280 IFACE_STAT(tx_errors, "tx_errors") \
2281 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2282 IFACE_STAT(rx_over_errors, "rx_over_err") \
2283 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2284 IFACE_STAT(collisions, "collisions")
2286 #define IFACE_STAT(MEMBER, NAME) + 1
2287 enum { N_IFACE_STATS = IFACE_STATS };
2289 int64_t values[N_IFACE_STATS];
2290 char *keys[N_IFACE_STATS];
2293 struct netdev_stats stats;
2295 if (iface_is_synthetic(iface)) {
2299 /* Intentionally ignore return value, since errors will set 'stats' to
2300 * all-1s, and we will deal with that correctly below. */
2301 netdev_get_stats(iface->netdev, &stats);
2303 /* Copy statistics into keys[] and values[]. */
2305 #define IFACE_STAT(MEMBER, NAME) \
2306 if (stats.MEMBER != UINT64_MAX) { \
2308 values[n] = stats.MEMBER; \
2313 ovs_assert(n <= N_IFACE_STATS);
2315 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2320 br_refresh_stp_status(struct bridge *br)
2322 struct smap smap = SMAP_INITIALIZER(&smap);
2323 struct ofproto *ofproto = br->ofproto;
2324 struct ofproto_stp_status status;
2326 if (ofproto_get_stp_status(ofproto, &status)) {
2330 if (!status.enabled) {
2331 ovsrec_bridge_set_status(br->cfg, NULL);
2335 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2336 STP_ID_ARGS(status.bridge_id));
2337 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2338 STP_ID_ARGS(status.designated_root));
2339 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2341 ovsrec_bridge_set_status(br->cfg, &smap);
2342 smap_destroy(&smap);
2346 port_refresh_stp_status(struct port *port)
2348 struct ofproto *ofproto = port->bridge->ofproto;
2349 struct iface *iface;
2350 struct ofproto_port_stp_status status;
2353 if (port_is_synthetic(port)) {
2357 /* STP doesn't currently support bonds. */
2358 if (!list_is_singleton(&port->ifaces)) {
2359 ovsrec_port_set_status(port->cfg, NULL);
2363 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2364 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2368 if (!status.enabled) {
2369 ovsrec_port_set_status(port->cfg, NULL);
2373 /* Set Status column. */
2375 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2376 smap_add(&smap, "stp_state", stp_state_name(status.state));
2377 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2378 smap_add(&smap, "stp_role", stp_role_name(status.role));
2379 ovsrec_port_set_status(port->cfg, &smap);
2380 smap_destroy(&smap);
2384 port_refresh_stp_stats(struct port *port)
2386 struct ofproto *ofproto = port->bridge->ofproto;
2387 struct iface *iface;
2388 struct ofproto_port_stp_stats stats;
2390 int64_t int_values[3];
2392 if (port_is_synthetic(port)) {
2396 /* STP doesn't currently support bonds. */
2397 if (!list_is_singleton(&port->ifaces)) {
2401 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2402 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2406 if (!stats.enabled) {
2407 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2411 /* Set Statistics column. */
2412 keys[0] = "stp_tx_count";
2413 int_values[0] = stats.tx_count;
2414 keys[1] = "stp_rx_count";
2415 int_values[1] = stats.rx_count;
2416 keys[2] = "stp_error_count";
2417 int_values[2] = stats.error_count;
2419 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2420 ARRAY_SIZE(int_values));
2424 br_refresh_rstp_status(struct bridge *br)
2426 struct smap smap = SMAP_INITIALIZER(&smap);
2427 struct ofproto *ofproto = br->ofproto;
2428 struct ofproto_rstp_status status;
2430 if (ofproto_get_rstp_status(ofproto, &status)) {
2433 if (!status.enabled) {
2434 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2437 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2438 RSTP_ID_ARGS(status.bridge_id));
2439 smap_add_format(&smap, "rstp_root_path_cost", "%d",
2440 status.root_path_cost);
2441 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2442 RSTP_ID_ARGS(status.root_id));
2443 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2444 RSTP_ID_ARGS(status.designated_id));
2445 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2446 status.designated_port_id);
2447 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2448 status.bridge_port_id);
2449 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2450 smap_destroy(&smap);
2454 port_refresh_rstp_status(struct port *port)
2456 struct ofproto *ofproto = port->bridge->ofproto;
2457 struct iface *iface;
2458 struct ofproto_port_rstp_status status;
2460 int64_t int_values[3];
2463 if (port_is_synthetic(port)) {
2467 /* RSTP doesn't currently support bonds. */
2468 if (!list_is_singleton(&port->ifaces)) {
2469 ovsrec_port_set_rstp_status(port->cfg, NULL);
2473 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2474 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2478 if (!status.enabled) {
2479 ovsrec_port_set_rstp_status(port->cfg, NULL);
2480 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2483 /* Set Status column. */
2486 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2488 smap_add_format(&smap, "rstp_port_role", "%s",
2489 rstp_port_role_name(status.role));
2490 smap_add_format(&smap, "rstp_port_state", "%s",
2491 rstp_state_name(status.state));
2493 ovsrec_port_set_rstp_status(port->cfg, &smap);
2494 smap_destroy(&smap);
2496 /* Set Statistics column. */
2497 keys[0] = "rstp_tx_count";
2498 int_values[0] = status.tx_count;
2499 keys[1] = "rstp_rx_count";
2500 int_values[1] = status.rx_count;
2501 keys[2] = "rstp_uptime";
2502 int_values[2] = status.uptime;
2503 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2504 ARRAY_SIZE(int_values));
2508 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2510 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2514 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2516 bool enable = enable_system_stats(cfg);
2518 system_stats_enable(enable);
2520 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2525 run_system_stats(void)
2527 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2530 stats = system_stats_run();
2532 struct ovsdb_idl_txn *txn;
2533 struct ovsdb_datum datum;
2535 txn = ovsdb_idl_txn_create(idl);
2536 ovsdb_datum_from_smap(&datum, stats);
2537 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2539 ovsdb_idl_txn_commit(txn);
2540 ovsdb_idl_txn_destroy(txn);
2547 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2550 case OFPCR12_ROLE_EQUAL:
2552 case OFPCR12_ROLE_MASTER:
2554 case OFPCR12_ROLE_SLAVE:
2556 case OFPCR12_ROLE_NOCHANGE:
2558 return "*** INVALID ROLE ***";
2563 refresh_controller_status(void)
2567 const struct ovsrec_controller *cfg;
2571 /* Accumulate status for controllers on all bridges. */
2572 HMAP_FOR_EACH (br, node, &all_bridges) {
2573 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2576 /* Update each controller in the database with current status. */
2577 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2578 struct ofproto_controller_info *cinfo =
2579 shash_find_data(&info, cfg->target);
2582 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2583 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2585 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2587 ovsrec_controller_set_is_connected(cfg, false);
2588 ovsrec_controller_set_role(cfg, NULL);
2589 ovsrec_controller_set_status(cfg, NULL);
2593 ofproto_free_ofproto_controller_info(&info);
2596 /* Update interface and mirror statistics if necessary. */
2598 run_stats_update(void)
2600 static struct ovsdb_idl_txn *stats_txn;
2601 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2608 /* Statistics update interval should always be greater than or equal to
2610 stats_interval = MAX(smap_get_int(&cfg->other_config,
2611 "stats-update-interval",
2613 if (stats_timer_interval != stats_interval) {
2614 stats_timer_interval = stats_interval;
2615 stats_timer = LLONG_MIN;
2618 if (time_msec() >= stats_timer) {
2619 enum ovsdb_idl_txn_status status;
2621 /* Rate limit the update. Do not start a new update if the
2622 * previous one is not done. */
2626 stats_txn = ovsdb_idl_txn_create(idl);
2627 HMAP_FOR_EACH (br, node, &all_bridges) {
2631 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2632 struct iface *iface;
2634 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2635 iface_refresh_stats(iface);
2637 port_refresh_stp_stats(port);
2639 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2640 mirror_refresh_stats(m);
2643 refresh_controller_status();
2646 status = ovsdb_idl_txn_commit(stats_txn);
2647 if (status != TXN_INCOMPLETE) {
2648 stats_timer = time_msec() + stats_timer_interval;
2649 ovsdb_idl_txn_destroy(stats_txn);
2655 /* Update bridge/port/interface status if necessary. */
2657 run_status_update(void)
2662 /* Rate limit the update. Do not start a new update if the
2663 * previous one is not done. */
2664 seq = seq_read(connectivity_seq_get());
2665 if (seq != connectivity_seqno || status_txn_try_again) {
2668 connectivity_seqno = seq;
2669 status_txn = ovsdb_idl_txn_create(idl);
2670 HMAP_FOR_EACH (br, node, &all_bridges) {
2673 br_refresh_stp_status(br);
2674 br_refresh_rstp_status(br);
2675 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2676 struct iface *iface;
2678 port_refresh_stp_status(port);
2679 port_refresh_rstp_status(port);
2680 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2681 iface_refresh_netdev_status(iface);
2682 iface_refresh_ofproto_status(iface);
2689 /* Commit the transaction and get the status. If the transaction finishes,
2690 * then destroy the transaction. Otherwise, keep it so that we can check
2691 * progress the next time that this function is called. */
2693 enum ovsdb_idl_txn_status status;
2695 status = ovsdb_idl_txn_commit(status_txn);
2696 if (status != TXN_INCOMPLETE) {
2697 ovsdb_idl_txn_destroy(status_txn);
2700 /* Sets the 'status_txn_try_again' if the transaction fails. */
2701 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2702 status_txn_try_again = false;
2704 status_txn_try_again = true;
2711 status_update_wait(void)
2713 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2714 * transaction to complete. If the status update to database needs to be
2715 * run again (transaction fails), registers a timeout in
2716 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2717 * sequence number. */
2719 ovsdb_idl_txn_wait(status_txn);
2720 } else if (status_txn_try_again) {
2721 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2723 seq_wait(connectivity_seq_get(), connectivity_seqno);
2734 /* Let each datapath type do the work that it needs to do. */
2736 ofproto_enumerate_types(&types);
2737 SSET_FOR_EACH (type, &types) {
2738 ofproto_type_run(type);
2740 sset_destroy(&types);
2742 /* Let each bridge do the work that it needs to do. */
2743 HMAP_FOR_EACH (br, node, &all_bridges) {
2744 ofproto_run(br->ofproto);
2751 static struct ovsrec_open_vswitch null_cfg;
2752 const struct ovsrec_open_vswitch *cfg;
2754 bool vlan_splinters_changed;
2756 ovsrec_open_vswitch_init(&null_cfg);
2760 if (ovsdb_idl_is_lock_contended(idl)) {
2761 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2762 struct bridge *br, *next_br;
2764 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2765 "disabling this process (pid %ld) until it goes away",
2766 (long int) getpid());
2768 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2771 /* Since we will not be running system_stats_run() in this process
2772 * with the current situation of multiple ovs-vswitchd daemons,
2773 * disable system stats collection. */
2774 system_stats_enable(false);
2775 /* This prevents the process from constantly waking up on
2776 * connectivity seq. */
2777 connectivity_seqno = seq_read(connectivity_seq_get());
2779 } else if (!ovsdb_idl_has_lock(idl)) {
2782 cfg = ovsrec_open_vswitch_first(idl);
2784 /* Initialize the ofproto library. This only needs to run once, but
2785 * it must be done after the configuration is set. If the
2786 * initialization has already occurred, bridge_init_ofproto()
2787 * returns immediately. */
2788 bridge_init_ofproto(cfg);
2790 /* Once the value of flow-restore-wait is false, we no longer should
2791 * check its value from the database. */
2792 if (cfg && ofproto_get_flow_restore_wait()) {
2793 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2794 "flow-restore-wait", false));
2799 /* Re-configure SSL. We do this on every trip through the main loop,
2800 * instead of just when the database changes, because the contents of the
2801 * key and certificate files can change without the database changing.
2803 * We do this before bridge_reconfigure() because that function might
2804 * initiate SSL connections and thus requires SSL to be configured. */
2805 if (cfg && cfg->ssl) {
2806 const struct ovsrec_ssl *ssl = cfg->ssl;
2808 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2809 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2812 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2813 * usage has changed. */
2814 vlan_splinters_changed = false;
2815 if (vlan_splinters_enabled_anywhere) {
2818 HMAP_FOR_EACH (br, node, &all_bridges) {
2819 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2820 vlan_splinters_changed = true;
2826 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2827 struct ovsdb_idl_txn *txn;
2829 idl_seqno = ovsdb_idl_get_seqno(idl);
2830 txn = ovsdb_idl_txn_create(idl);
2831 bridge_reconfigure(cfg ? cfg : &null_cfg);
2834 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2837 /* If we are completing our initial configuration for this run
2838 * of ovs-vswitchd, then keep the transaction around to monitor
2839 * it for completion. */
2840 if (initial_config_done) {
2841 /* Always sets the 'status_txn_try_again' to check again,
2842 * in case that this transaction fails. */
2843 status_txn_try_again = true;
2844 ovsdb_idl_txn_commit(txn);
2845 ovsdb_idl_txn_destroy(txn);
2847 initial_config_done = true;
2848 daemonize_txn = txn;
2852 if (daemonize_txn) {
2853 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2854 if (status != TXN_INCOMPLETE) {
2855 ovsdb_idl_txn_destroy(daemonize_txn);
2856 daemonize_txn = NULL;
2858 /* ovs-vswitchd has completed initialization, so allow the
2859 * process that forked us to exit successfully. */
2860 daemonize_complete();
2862 vlog_enable_async();
2864 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2869 run_status_update();
2879 ovsdb_idl_wait(idl);
2880 if (daemonize_txn) {
2881 ovsdb_idl_txn_wait(daemonize_txn);
2885 ofproto_enumerate_types(&types);
2886 SSET_FOR_EACH (type, &types) {
2887 ofproto_type_wait(type);
2889 sset_destroy(&types);
2891 if (!hmap_is_empty(&all_bridges)) {
2894 HMAP_FOR_EACH (br, node, &all_bridges) {
2895 ofproto_wait(br->ofproto);
2898 poll_timer_wait_until(stats_timer);
2901 status_update_wait();
2902 system_stats_wait();
2905 /* Adds some memory usage statistics for bridges into 'usage', for use with
2906 * memory_report(). */
2908 bridge_get_memory_usage(struct simap *usage)
2915 ofproto_enumerate_types(&types);
2916 SSET_FOR_EACH (type, &types) {
2917 ofproto_type_get_memory_usage(type, usage);
2919 sset_destroy(&types);
2921 HMAP_FOR_EACH (br, node, &all_bridges) {
2922 ofproto_get_memory_usage(br->ofproto, usage);
2926 /* QoS unixctl user interface functions. */
2928 struct qos_unixctl_show_cbdata {
2930 struct iface *iface;
2934 qos_unixctl_show_queue(unsigned int queue_id,
2935 const struct smap *details,
2936 struct iface *iface,
2939 struct netdev_queue_stats stats;
2940 struct smap_node *node;
2943 ds_put_cstr(ds, "\n");
2945 ds_put_format(ds, "Queue %u:\n", queue_id);
2947 ds_put_cstr(ds, "Default:\n");
2950 SMAP_FOR_EACH (node, details) {
2951 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2954 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2956 if (stats.tx_packets != UINT64_MAX) {
2957 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2960 if (stats.tx_bytes != UINT64_MAX) {
2961 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2964 if (stats.tx_errors != UINT64_MAX) {
2965 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2968 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2969 queue_id, ovs_strerror(error));
2974 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2975 const char *argv[], void *aux OVS_UNUSED)
2977 struct ds ds = DS_EMPTY_INITIALIZER;
2978 struct smap smap = SMAP_INITIALIZER(&smap);
2979 struct iface *iface;
2981 struct smap_node *node;
2983 iface = iface_find(argv[1]);
2985 unixctl_command_reply_error(conn, "no such interface");
2989 netdev_get_qos(iface->netdev, &type, &smap);
2991 if (*type != '\0') {
2992 struct netdev_queue_dump dump;
2993 struct smap details;
2994 unsigned int queue_id;
2996 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2998 SMAP_FOR_EACH (node, &smap) {
2999 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3002 smap_init(&details);
3003 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3004 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3006 smap_destroy(&details);
3008 unixctl_command_reply(conn, ds_cstr(&ds));
3010 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3011 unixctl_command_reply_error(conn, ds_cstr(&ds));
3014 smap_destroy(&smap);
3018 /* Bridge reconfiguration functions. */
3020 bridge_create(const struct ovsrec_bridge *br_cfg)
3024 ovs_assert(!bridge_lookup(br_cfg->name));
3025 br = xzalloc(sizeof *br);
3027 br->name = xstrdup(br_cfg->name);
3028 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3031 /* Derive the default Ethernet address from the bridge's UUID. This should
3032 * be unique and it will be stable between ovs-vswitchd runs. */
3033 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3034 eth_addr_mark_random(br->default_ea);
3036 hmap_init(&br->ports);
3037 hmap_init(&br->ifaces);
3038 hmap_init(&br->iface_by_name);
3039 hmap_init(&br->mirrors);
3041 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3045 bridge_destroy(struct bridge *br)
3048 struct mirror *mirror, *next_mirror;
3049 struct port *port, *next_port;
3051 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3054 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3055 mirror_destroy(mirror);
3058 hmap_remove(&all_bridges, &br->node);
3059 ofproto_destroy(br->ofproto);
3060 hmap_destroy(&br->ifaces);
3061 hmap_destroy(&br->ports);
3062 hmap_destroy(&br->iface_by_name);
3063 hmap_destroy(&br->mirrors);
3070 static struct bridge *
3071 bridge_lookup(const char *name)
3075 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3076 if (!strcmp(br->name, name)) {
3083 /* Handle requests for a listing of all flows known by the OpenFlow
3084 * stack, including those normally hidden. */
3086 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3087 const char *argv[], void *aux OVS_UNUSED)
3092 br = bridge_lookup(argv[1]);
3094 unixctl_command_reply_error(conn, "Unknown bridge");
3099 ofproto_get_all_flows(br->ofproto, &results);
3101 unixctl_command_reply(conn, ds_cstr(&results));
3102 ds_destroy(&results);
3105 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3106 * connections and reconnect. If BRIDGE is not specified, then all bridges
3107 * drop their controller connections and reconnect. */
3109 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3110 const char *argv[], void *aux OVS_UNUSED)
3114 br = bridge_lookup(argv[1]);
3116 unixctl_command_reply_error(conn, "Unknown bridge");
3119 ofproto_reconnect_controllers(br->ofproto);
3121 HMAP_FOR_EACH (br, node, &all_bridges) {
3122 ofproto_reconnect_controllers(br->ofproto);
3125 unixctl_command_reply(conn, NULL);
3129 bridge_get_controllers(const struct bridge *br,
3130 struct ovsrec_controller ***controllersp)
3132 struct ovsrec_controller **controllers;
3133 size_t n_controllers;
3135 controllers = br->cfg->controller;
3136 n_controllers = br->cfg->n_controller;
3138 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3144 *controllersp = controllers;
3146 return n_controllers;
3150 bridge_collect_wanted_ports(struct bridge *br,
3151 const unsigned long int *splinter_vlans,
3152 struct shash *wanted_ports)
3156 shash_init(wanted_ports);
3158 for (i = 0; i < br->cfg->n_ports; i++) {
3159 const char *name = br->cfg->ports[i]->name;
3160 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3161 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3165 if (bridge_get_controllers(br, NULL)
3166 && !shash_find(wanted_ports, br->name)) {
3167 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3168 br->name, br->name);
3170 ovsrec_interface_init(&br->synth_local_iface);
3171 ovsrec_port_init(&br->synth_local_port);
3173 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3174 br->synth_local_port.n_interfaces = 1;
3175 br->synth_local_port.name = br->name;
3177 br->synth_local_iface.name = br->name;
3178 br->synth_local_iface.type = "internal";
3180 br->synth_local_ifacep = &br->synth_local_iface;
3182 shash_add(wanted_ports, br->name, &br->synth_local_port);
3185 if (splinter_vlans) {
3186 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3190 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3191 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3192 * 'br' needs to complete its configuration. */
3194 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3196 struct shash_node *port_node;
3197 struct port *port, *next;
3199 /* Get rid of deleted ports.
3200 * Get rid of deleted interfaces on ports that still exist. */
3201 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3202 port->cfg = shash_find_data(wanted_ports, port->name);
3206 port_del_ifaces(port);
3210 /* Update iface->cfg and iface->type in interfaces that still exist. */
3211 SHASH_FOR_EACH (port_node, wanted_ports) {
3212 const struct ovsrec_port *port = port_node->data;
3215 for (i = 0; i < port->n_interfaces; i++) {
3216 const struct ovsrec_interface *cfg = port->interfaces[i];
3217 struct iface *iface = iface_lookup(br, cfg->name);
3218 const char *type = iface_get_type(cfg, br->cfg);
3223 } else if (!strcmp(type, "null")) {
3224 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3225 " may be removed in February 2013. Please email"
3226 " dev@openvswitch.org with concerns.",
3229 /* We will add new interfaces later. */
3235 /* Initializes 'oc' appropriately as a management service controller for
3238 * The caller must free oc->target when it is no longer needed. */
3240 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3241 struct ofproto_controller *oc)
3243 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3244 oc->max_backoff = 0;
3245 oc->probe_interval = 60;
3246 oc->band = OFPROTO_OUT_OF_BAND;
3248 oc->burst_limit = 0;
3249 oc->enable_async_msgs = true;
3253 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3255 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3256 struct ofproto_controller *oc)
3260 oc->target = c->target;
3261 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3262 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3263 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3264 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3265 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3266 oc->burst_limit = (c->controller_burst_limit
3267 ? *c->controller_burst_limit : 0);
3268 oc->enable_async_msgs = (!c->enable_async_messages
3269 || *c->enable_async_messages);
3270 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3271 if (dscp < 0 || dscp > 63) {
3272 dscp = DSCP_DEFAULT;
3277 /* Configures the IP stack for 'br''s local interface properly according to the
3278 * configuration in 'c'. */
3280 bridge_configure_local_iface_netdev(struct bridge *br,
3281 struct ovsrec_controller *c)
3283 struct netdev *netdev;
3284 struct in_addr mask, gateway;
3286 struct iface *local_iface;
3289 /* If there's no local interface or no IP address, give up. */
3290 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3291 if (!local_iface || !c->local_ip
3292 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3296 /* Bring up the local interface. */
3297 netdev = local_iface->netdev;
3298 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3300 /* Configure the IP address and netmask. */
3301 if (!c->local_netmask
3302 || !inet_pton(AF_INET, c->local_netmask, &mask)
3304 mask.s_addr = guess_netmask(ip.s_addr);
3306 if (!netdev_set_in4(netdev, ip, mask)) {
3307 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3308 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3311 /* Configure the default gateway. */
3312 if (c->local_gateway
3313 && inet_pton(AF_INET, c->local_gateway, &gateway)
3314 && gateway.s_addr) {
3315 if (!netdev_add_router(netdev, gateway)) {
3316 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3317 br->name, IP_ARGS(gateway.s_addr));
3322 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3323 * in either string are treated as equal to any number of slashes in the other,
3324 * e.g. "x///y" is equal to "x/y".
3326 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3327 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3328 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3329 * 'b' against a prefix of 'a'.
3332 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3334 const char *b_start = b;
3336 if (b - b_start >= b_stoplen) {
3338 } else if (*a != *b) {
3340 } else if (*a == '/') {
3341 a += strspn(a, "/");
3342 b += strspn(b, "/");
3343 } else if (*a == '\0') {
3353 bridge_configure_remotes(struct bridge *br,
3354 const struct sockaddr_in *managers, size_t n_managers)
3356 bool disable_in_band;
3358 struct ovsrec_controller **controllers;
3359 size_t n_controllers;
3361 enum ofproto_fail_mode fail_mode;
3363 struct ofproto_controller *ocs;
3367 /* Check if we should disable in-band control on this bridge. */
3368 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3371 /* Set OpenFlow queue ID for in-band control. */
3372 ofproto_set_in_band_queue(br->ofproto,
3373 smap_get_int(&br->cfg->other_config,
3374 "in-band-queue", -1));
3376 if (disable_in_band) {
3377 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3379 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3382 n_controllers = bridge_get_controllers(br, &controllers);
3384 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3387 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3388 for (i = 0; i < n_controllers; i++) {
3389 struct ovsrec_controller *c = controllers[i];
3391 if (!strncmp(c->target, "punix:", 6)
3392 || !strncmp(c->target, "unix:", 5)) {
3393 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3396 if (!strncmp(c->target, "unix:", 5)) {
3397 /* Connect to a listening socket */
3398 whitelist = xasprintf("unix:%s/", ovs_rundir());
3399 if (strchr(c->target, '/') &&
3400 !equal_pathnames(c->target, whitelist,
3401 strlen(whitelist))) {
3402 /* Absolute path specified, but not in ovs_rundir */
3403 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3404 "controller \"%s\" due to possibility for "
3405 "remote exploit. Instead, specify socket "
3406 "in whitelisted \"%s\" or connect to "
3407 "\"unix:%s/%s.mgmt\" (which is always "
3408 "available without special configuration).",
3409 br->name, c->target, whitelist,
3410 ovs_rundir(), br->name);
3415 whitelist = xasprintf("punix:%s/%s.controller",
3416 ovs_rundir(), br->name);
3417 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3418 /* Prevent remote ovsdb-server users from accessing
3419 * arbitrary Unix domain sockets and overwriting arbitrary
3421 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3422 "controller \"%s\" due to possibility of "
3423 "overwriting local files. Instead, specify "
3424 "whitelisted \"%s\" or connect to "
3425 "\"unix:%s/%s.mgmt\" (which is always "
3426 "available without special configuration).",
3427 br->name, c->target, whitelist,
3428 ovs_rundir(), br->name);
3437 bridge_configure_local_iface_netdev(br, c);
3438 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3439 if (disable_in_band) {
3440 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3445 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3446 bridge_get_allowed_versions(br));
3447 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3450 /* Set the fail-mode. */
3451 fail_mode = !br->cfg->fail_mode
3452 || !strcmp(br->cfg->fail_mode, "standalone")
3453 ? OFPROTO_FAIL_STANDALONE
3454 : OFPROTO_FAIL_SECURE;
3455 ofproto_set_fail_mode(br->ofproto, fail_mode);
3457 /* Configure OpenFlow controller connection snooping. */
3458 if (!ofproto_has_snoops(br->ofproto)) {
3462 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3463 ovs_rundir(), br->name));
3464 ofproto_set_snoops(br->ofproto, &snoops);
3465 sset_destroy(&snoops);
3470 bridge_configure_tables(struct bridge *br)
3472 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3476 n_tables = ofproto_get_n_tables(br->ofproto);
3478 for (i = 0; i < n_tables; i++) {
3479 struct ofproto_table_settings s;
3480 bool use_default_prefixes = true;
3483 s.max_flows = UINT_MAX;
3486 s.n_prefix_fields = 0;
3487 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3489 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3490 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3493 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3494 s.max_flows = *cfg->flow_limit;
3496 if (cfg->overflow_policy
3497 && !strcmp(cfg->overflow_policy, "evict")) {
3499 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3500 for (k = 0; k < cfg->n_groups; k++) {
3501 const char *string = cfg->groups[k];
3504 msg = mf_parse_subfield__(&s.groups[k], &string);
3506 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3507 "'groups' (%s)", br->name, i, msg);
3509 } else if (*string) {
3510 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3511 "element '%s' contains trailing garbage",
3512 br->name, i, cfg->groups[k]);
3518 /* Prefix lookup fields. */
3519 s.n_prefix_fields = 0;
3520 for (k = 0; k < cfg->n_prefixes; k++) {
3521 const char *name = cfg->prefixes[k];
3522 const struct mf_field *mf;
3524 if (strcmp(name, "none") == 0) {
3525 use_default_prefixes = false;
3526 s.n_prefix_fields = 0;
3529 mf = mf_from_name(name);
3531 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3535 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3536 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3537 "%s", br->name, name);
3540 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3541 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3542 "field not used: %s", br->name, name);
3545 use_default_prefixes = false;
3546 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3549 if (use_default_prefixes) {
3550 /* Use default values. */
3551 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3552 memcpy(s.prefix_fields, default_prefix_fields,
3553 sizeof default_prefix_fields);
3556 struct ds ds = DS_EMPTY_INITIALIZER;
3557 for (k = 0; k < s.n_prefix_fields; k++) {
3559 ds_put_char(&ds, ',');
3561 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3563 if (s.n_prefix_fields == 0) {
3564 ds_put_cstr(&ds, "none");
3566 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3567 br->name, i, ds_cstr(&ds));
3571 ofproto_configure_table(br->ofproto, i, &s);
3575 for (; j < br->cfg->n_flow_tables; j++) {
3576 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3577 "%"PRId64" not supported by this datapath", br->name,
3578 br->cfg->key_flow_tables[j]);
3583 bridge_configure_dp_desc(struct bridge *br)
3585 ofproto_set_dp_desc(br->ofproto,
3586 smap_get(&br->cfg->other_config, "dp-desc"));
3589 /* Port functions. */
3591 static struct port *
3592 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3596 port = xzalloc(sizeof *port);
3598 port->name = xstrdup(cfg->name);
3600 list_init(&port->ifaces);
3602 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3606 /* Deletes interfaces from 'port' that are no longer configured for it. */
3608 port_del_ifaces(struct port *port)
3610 struct iface *iface, *next;
3611 struct sset new_ifaces;
3614 /* Collect list of new interfaces. */
3615 sset_init(&new_ifaces);
3616 for (i = 0; i < port->cfg->n_interfaces; i++) {
3617 const char *name = port->cfg->interfaces[i]->name;
3618 const char *type = port->cfg->interfaces[i]->type;
3619 if (strcmp(type, "null")) {
3620 sset_add(&new_ifaces, name);
3624 /* Get rid of deleted interfaces. */
3625 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3626 if (!sset_contains(&new_ifaces, iface->name)) {
3627 iface_destroy(iface);
3631 sset_destroy(&new_ifaces);
3635 port_destroy(struct port *port)
3638 struct bridge *br = port->bridge;
3639 struct iface *iface, *next;
3642 ofproto_bundle_unregister(br->ofproto, port);
3645 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3646 iface_destroy__(iface);
3649 hmap_remove(&br->ports, &port->hmap_node);
3655 static struct port *
3656 port_lookup(const struct bridge *br, const char *name)
3660 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3662 if (!strcmp(port->name, name)) {
3670 enable_lacp(struct port *port, bool *activep)
3672 if (!port->cfg->lacp) {
3673 /* XXX when LACP implementation has been sufficiently tested, enable by
3674 * default and make active on bonded ports. */
3676 } else if (!strcmp(port->cfg->lacp, "off")) {
3678 } else if (!strcmp(port->cfg->lacp, "active")) {
3681 } else if (!strcmp(port->cfg->lacp, "passive")) {
3685 VLOG_WARN("port %s: unknown LACP mode %s",
3686 port->name, port->cfg->lacp);
3691 static struct lacp_settings *
3692 port_configure_lacp(struct port *port, struct lacp_settings *s)
3694 const char *lacp_time, *system_id;
3697 if (!enable_lacp(port, &s->active)) {
3701 s->name = port->name;
3703 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3705 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3706 ETH_ADDR_SCAN_ARGS(s->id))) {
3707 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3708 " address.", port->name, system_id);
3712 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3715 if (eth_addr_is_zero(s->id)) {
3716 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3720 /* Prefer bondable links if unspecified. */
3721 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3723 s->priority = (priority > 0 && priority <= UINT16_MAX
3725 : UINT16_MAX - !list_is_short(&port->ifaces));
3727 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3728 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3730 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3731 "lacp-fallback-ab", false);
3737 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3739 int priority, portid, key;
3741 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3742 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3744 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3746 if (portid <= 0 || portid > UINT16_MAX) {
3747 portid = ofp_to_u16(iface->ofp_port);
3750 if (priority <= 0 || priority > UINT16_MAX) {
3751 priority = UINT16_MAX;
3754 if (key < 0 || key > UINT16_MAX) {
3758 s->name = iface->name;
3760 s->priority = priority;
3765 port_configure_bond(struct port *port, struct bond_settings *s)
3767 const char *detect_s;
3768 struct iface *iface;
3769 int miimon_interval;
3771 s->name = port->name;
3773 if (port->cfg->bond_mode) {
3774 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3775 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3776 port->name, port->cfg->bond_mode,
3777 bond_mode_to_string(s->balance));
3780 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3782 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3783 * active-backup. At some point we should remove this warning. */
3784 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3785 " in previous versions, the default bond_mode was"
3786 " balance-slb", port->name,
3787 bond_mode_to_string(s->balance));
3789 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3790 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3791 "please use another bond type or disable flood_vlans",
3795 miimon_interval = smap_get_int(&port->cfg->other_config,
3796 "bond-miimon-interval", 0);
3797 if (miimon_interval <= 0) {
3798 miimon_interval = 200;
3801 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3802 if (!detect_s || !strcmp(detect_s, "carrier")) {
3803 miimon_interval = 0;
3804 } else if (strcmp(detect_s, "miimon")) {
3805 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3806 "defaulting to carrier", port->name, detect_s);
3807 miimon_interval = 0;
3810 s->up_delay = MAX(0, port->cfg->bond_updelay);
3811 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3812 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3813 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3814 "bond-rebalance-interval", 10000);
3815 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3816 s->rebalance_interval = 1000;
3819 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3820 "lacp-fallback-ab", false);
3822 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3823 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3827 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3828 * instead of obtaining it from the database. */
3830 port_is_synthetic(const struct port *port)
3832 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3835 /* Interface functions. */
3838 iface_is_internal(const struct ovsrec_interface *iface,
3839 const struct ovsrec_bridge *br)
3841 /* The local port and "internal" ports are always "internal". */
3842 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3845 /* Returns the correct network device type for interface 'iface' in bridge
3848 iface_get_type(const struct ovsrec_interface *iface,
3849 const struct ovsrec_bridge *br)
3853 /* The local port always has type "internal". Other ports take
3854 * their type from the database and default to "system" if none is
3856 if (iface_is_internal(iface, br)) {
3859 type = iface->type[0] ? iface->type : "system";
3862 return ofproto_port_open_type(br->datapath_type, type);
3866 iface_destroy__(struct iface *iface)
3869 struct port *port = iface->port;
3870 struct bridge *br = port->bridge;
3872 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3873 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3876 if (iface->ofp_port != OFPP_NONE) {
3877 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3880 list_remove(&iface->port_elem);
3881 hmap_remove(&br->iface_by_name, &iface->name_node);
3883 /* The user is changing configuration here, so netdev_remove needs to be
3884 * used as opposed to netdev_close */
3885 netdev_remove(iface->netdev);
3893 iface_destroy(struct iface *iface)
3896 struct port *port = iface->port;
3898 iface_destroy__(iface);
3899 if (list_is_empty(&port->ifaces)) {
3905 static struct iface *
3906 iface_lookup(const struct bridge *br, const char *name)
3908 struct iface *iface;
3910 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3911 &br->iface_by_name) {
3912 if (!strcmp(iface->name, name)) {
3920 static struct iface *
3921 iface_find(const char *name)
3923 const struct bridge *br;
3925 HMAP_FOR_EACH (br, node, &all_bridges) {
3926 struct iface *iface = iface_lookup(br, name);
3935 static struct iface *
3936 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3938 struct iface *iface;
3940 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3942 if (iface->ofp_port == ofp_port) {
3949 /* Set Ethernet address of 'iface', if one is specified in the configuration
3952 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3954 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3955 struct iface *hw_addr_iface;
3957 if (strcmp(iface->type, "internal")) {
3961 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3963 } else if (port->cfg->fake_bridge) {
3964 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3965 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3970 if (iface->ofp_port == OFPP_LOCAL) {
3971 VLOG_ERR("interface %s: ignoring mac in Interface record "
3972 "(use Bridge record to set local port's mac)",
3974 } else if (eth_addr_is_multicast(mac)) {
3975 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3978 int error = netdev_set_etheraddr(iface->netdev, mac);
3980 VLOG_ERR("interface %s: setting MAC failed (%s)",
3981 iface->name, ovs_strerror(error));
3987 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3989 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3991 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3992 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3993 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3997 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3998 * sets the "ofport" field to -1.
4000 * This is appropriate when 'if_cfg''s interface cannot be created or is
4001 * otherwise invalid. */
4003 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4005 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4006 iface_set_ofport(if_cfg, OFPP_NONE);
4007 ovsrec_interface_set_error(if_cfg, errp);
4008 ovsrec_interface_set_status(if_cfg, NULL);
4009 ovsrec_interface_set_admin_state(if_cfg, NULL);
4010 ovsrec_interface_set_duplex(if_cfg, NULL);
4011 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4012 ovsrec_interface_set_link_state(if_cfg, NULL);
4013 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4014 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4015 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4016 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4017 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4018 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4019 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4020 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4025 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4027 union ovsdb_atom atom;
4029 atom.integer = target;
4030 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4034 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4036 struct ofpbuf queues_buf;
4038 ofpbuf_init(&queues_buf, 0);
4040 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
4041 netdev_set_qos(iface->netdev, NULL, NULL);
4043 const struct ovsdb_datum *queues;
4044 struct netdev_queue_dump dump;
4045 unsigned int queue_id;
4046 struct smap details;
4050 /* Configure top-level Qos for 'iface'. */
4051 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4053 /* Deconfigure queues that were deleted. */
4054 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4056 smap_init(&details);
4057 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4058 if (!queue_ids_include(queues, queue_id)) {
4059 netdev_delete_queue(iface->netdev, queue_id);
4062 smap_destroy(&details);
4064 /* Configure queues for 'iface'. */
4066 for (i = 0; i < qos->n_queues; i++) {
4067 const struct ovsrec_queue *queue = qos->value_queues[i];
4068 unsigned int queue_id = qos->key_queues[i];
4070 if (queue_id == 0) {
4074 if (queue->n_dscp == 1) {
4075 struct ofproto_port_queue *port_queue;
4077 port_queue = ofpbuf_put_uninit(&queues_buf,
4078 sizeof *port_queue);
4079 port_queue->queue = queue_id;
4080 port_queue->dscp = queue->dscp[0];
4083 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4086 struct smap details;
4088 smap_init(&details);
4089 netdev_set_queue(iface->netdev, 0, &details);
4090 smap_destroy(&details);
4094 if (iface->ofp_port != OFPP_NONE) {
4095 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
4096 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
4098 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4099 port_queues, n_queues);
4102 netdev_set_policing(iface->netdev,
4103 iface->cfg->ingress_policing_rate,
4104 iface->cfg->ingress_policing_burst);
4106 ofpbuf_uninit(&queues_buf);
4110 iface_configure_cfm(struct iface *iface)
4112 const struct ovsrec_interface *cfg = iface->cfg;
4113 const char *opstate_str;
4114 const char *cfm_ccm_vlan;
4115 struct cfm_settings s;
4116 struct smap netdev_args;
4118 if (!cfg->n_cfm_mpid) {
4119 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4123 s.check_tnl_key = false;
4124 smap_init(&netdev_args);
4125 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4126 const char *key = smap_get(&netdev_args, "key");
4127 const char *in_key = smap_get(&netdev_args, "in_key");
4129 s.check_tnl_key = (key && !strcmp(key, "flow"))
4130 || (in_key && !strcmp(in_key, "flow"));
4132 smap_destroy(&netdev_args);
4134 s.mpid = *cfg->cfm_mpid;
4135 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4136 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4137 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4139 if (s.interval <= 0) {
4143 if (!cfm_ccm_vlan) {
4145 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4146 s.ccm_vlan = CFM_RANDOM_VLAN;
4148 s.ccm_vlan = atoi(cfm_ccm_vlan);
4149 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4154 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4156 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4158 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4159 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4161 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4164 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4165 * instead of obtaining it from the database. */
4167 iface_is_synthetic(const struct iface *iface)
4169 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4173 iface_validate_ofport__(size_t n, int64_t *ofport)
4175 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4176 ? u16_to_ofp(*ofport)
4181 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4183 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4187 iface_pick_ofport(const struct ovsrec_interface *cfg)
4189 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4190 return (requested_ofport != OFPP_NONE
4192 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4195 /* Port mirroring. */
4197 static struct mirror *
4198 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4202 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4203 if (uuid_equals(uuid, &m->uuid)) {
4211 bridge_configure_mirrors(struct bridge *br)
4213 const struct ovsdb_datum *mc;
4214 unsigned long *flood_vlans;
4215 struct mirror *m, *next;
4218 /* Get rid of deleted mirrors. */
4219 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4220 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4221 union ovsdb_atom atom;
4223 atom.uuid = m->uuid;
4224 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4229 /* Add new mirrors and reconfigure existing ones. */
4230 for (i = 0; i < br->cfg->n_mirrors; i++) {
4231 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4232 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4234 m = mirror_create(br, cfg);
4237 if (!mirror_configure(m)) {
4242 /* Update flooded vlans (for RSPAN). */
4243 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4244 br->cfg->n_flood_vlans);
4245 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4246 bitmap_free(flood_vlans);
4249 static struct mirror *
4250 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4254 m = xzalloc(sizeof *m);
4255 m->uuid = cfg->header_.uuid;
4256 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4258 m->name = xstrdup(cfg->name);
4264 mirror_destroy(struct mirror *m)
4267 struct bridge *br = m->bridge;
4270 ofproto_mirror_unregister(br->ofproto, m);
4273 hmap_remove(&br->mirrors, &m->hmap_node);
4280 mirror_collect_ports(struct mirror *m,
4281 struct ovsrec_port **in_ports, int n_in_ports,
4282 void ***out_portsp, size_t *n_out_portsp)
4284 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4285 size_t n_out_ports = 0;
4288 for (i = 0; i < n_in_ports; i++) {
4289 const char *name = in_ports[i]->name;
4290 struct port *port = port_lookup(m->bridge, name);
4292 out_ports[n_out_ports++] = port;
4294 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4295 "port %s", m->bridge->name, m->name, name);
4298 *out_portsp = out_ports;
4299 *n_out_portsp = n_out_ports;
4303 mirror_configure(struct mirror *m)
4305 const struct ovsrec_mirror *cfg = m->cfg;
4306 struct ofproto_mirror_settings s;
4309 if (strcmp(cfg->name, m->name)) {
4311 m->name = xstrdup(cfg->name);
4315 /* Get output port or VLAN. */
4316 if (cfg->output_port) {
4317 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4318 if (!s.out_bundle) {
4319 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4320 m->bridge->name, m->name);
4323 s.out_vlan = UINT16_MAX;
4325 if (cfg->output_vlan) {
4326 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4327 "output vlan; ignoring output vlan",
4328 m->bridge->name, m->name);
4330 } else if (cfg->output_vlan) {
4331 /* The database should prevent invalid VLAN values. */
4332 s.out_bundle = NULL;
4333 s.out_vlan = *cfg->output_vlan;
4335 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4336 m->bridge->name, m->name);
4340 /* Get port selection. */
4341 if (cfg->select_all) {
4342 size_t n_ports = hmap_count(&m->bridge->ports);
4343 void **ports = xmalloc(n_ports * sizeof *ports);
4348 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4358 /* Get ports, dropping ports that don't exist.
4359 * The IDL ensures that there are no duplicates. */
4360 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4361 &s.srcs, &s.n_srcs);
4362 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4363 &s.dsts, &s.n_dsts);
4366 /* Get VLAN selection. */
4367 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4370 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4373 if (s.srcs != s.dsts) {
4382 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4384 * This is deprecated. It is only for compatibility with broken device drivers
4385 * in old versions of Linux that do not properly support VLANs when VLAN
4386 * devices are not used. When broken device drivers are no longer in
4387 * widespread use, we will delete these interfaces. */
4389 static struct ovsrec_port **recs;
4390 static size_t n_recs, allocated_recs;
4392 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4393 * splinters are reconfigured. */
4395 register_rec(struct ovsrec_port *rec)
4397 if (n_recs >= allocated_recs) {
4398 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4400 recs[n_recs++] = rec;
4403 /* Frees all of the ports registered with register_reg(). */
4405 free_registered_recs(void)
4409 for (i = 0; i < n_recs; i++) {
4410 struct ovsrec_port *port = recs[i];
4413 for (j = 0; j < port->n_interfaces; j++) {
4414 struct ovsrec_interface *iface = port->interfaces[j];
4419 smap_destroy(&port->other_config);
4420 free(port->interfaces);
4428 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4431 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4433 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4437 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4440 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4441 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4442 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4445 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4446 * use, returns NULL.
4448 * Updates 'vlan_splinters_enabled_anywhere'. */
4449 static unsigned long int *
4450 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4452 unsigned long int *splinter_vlans;
4453 struct sset splinter_ifaces;
4454 const char *real_dev_name;
4455 struct shash *real_devs;
4456 struct shash_node *node;
4460 /* Free space allocated for synthesized ports and interfaces, since we're
4461 * in the process of reconstructing all of them. */
4462 free_registered_recs();
4464 splinter_vlans = bitmap_allocate(4096);
4465 sset_init(&splinter_ifaces);
4466 vlan_splinters_enabled_anywhere = false;
4467 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4468 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4471 for (j = 0; j < br_cfg->n_ports; j++) {
4472 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4475 for (k = 0; k < port_cfg->n_interfaces; k++) {
4476 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4478 if (vlan_splinters_is_enabled(iface_cfg)) {
4479 vlan_splinters_enabled_anywhere = true;
4480 sset_add(&splinter_ifaces, iface_cfg->name);
4481 vlan_bitmap_from_array__(port_cfg->trunks,
4487 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4488 bitmap_set1(splinter_vlans, *port_cfg->tag);
4493 if (!vlan_splinters_enabled_anywhere) {
4494 free(splinter_vlans);
4495 sset_destroy(&splinter_ifaces);
4499 HMAP_FOR_EACH (br, node, &all_bridges) {
4501 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4505 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4506 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4507 * device to be created for it. */
4508 bitmap_set0(splinter_vlans, 0);
4509 bitmap_set0(splinter_vlans, 4095);
4511 /* Delete all VLAN devices that we don't need. */
4513 real_devs = vlandev_get_real_devs();
4514 SHASH_FOR_EACH (node, real_devs) {
4515 const struct vlan_real_dev *real_dev = node->data;
4516 const struct vlan_dev *vlan_dev;
4517 bool real_dev_has_splinters;
4519 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4521 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4522 if (!real_dev_has_splinters
4523 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4524 struct netdev *netdev;
4526 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4527 if (!netdev_get_in4(netdev, NULL, NULL) ||
4528 !netdev_get_in6(netdev, NULL)) {
4529 /* It has an IP address configured, so we don't own
4530 * it. Don't delete it. */
4532 vlandev_del(vlan_dev->name);
4534 netdev_close(netdev);
4541 /* Add all VLAN devices that we need. */
4542 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4545 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4546 if (!vlandev_get_name(real_dev_name, vid)) {
4547 vlandev_add(real_dev_name, vid);
4554 sset_destroy(&splinter_ifaces);
4556 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4557 free(splinter_vlans);
4560 return splinter_vlans;
4563 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4566 configure_splinter_port(struct port *port)
4568 struct ofproto *ofproto = port->bridge->ofproto;
4569 ofp_port_t realdev_ofp_port;
4570 const char *realdev_name;
4571 struct iface *vlandev, *realdev;
4573 ofproto_bundle_unregister(port->bridge->ofproto, port);
4575 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4578 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4579 realdev = iface_lookup(port->bridge, realdev_name);
4580 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4582 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4586 static struct ovsrec_port *
4587 synthesize_splinter_port(const char *real_dev_name,
4588 const char *vlan_dev_name, int vid)
4590 struct ovsrec_interface *iface;
4591 struct ovsrec_port *port;
4593 iface = xmalloc(sizeof *iface);
4594 ovsrec_interface_init(iface);
4595 iface->name = xstrdup(vlan_dev_name);
4596 iface->type = "system";
4598 port = xmalloc(sizeof *port);
4599 ovsrec_port_init(port);
4600 port->interfaces = xmemdup(&iface, sizeof iface);
4601 port->n_interfaces = 1;
4602 port->name = xstrdup(vlan_dev_name);
4603 port->vlan_mode = "splinter";
4604 port->tag = xmalloc(sizeof *port->tag);
4607 smap_add(&port->other_config, "realdev", real_dev_name);
4613 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4614 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4615 * 1-bit in the 'splinter_vlans' bitmap. */
4617 add_vlan_splinter_ports(struct bridge *br,
4618 const unsigned long int *splinter_vlans,
4619 struct shash *ports)
4623 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4624 * we're modifying 'ports'. */
4625 for (i = 0; i < br->cfg->n_ports; i++) {
4626 const char *name = br->cfg->ports[i]->name;
4627 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4630 for (j = 0; j < port_cfg->n_interfaces; j++) {
4631 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4633 if (vlan_splinters_is_enabled(iface_cfg)) {
4634 const char *real_dev_name;
4637 real_dev_name = iface_cfg->name;
4638 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4639 const char *vlan_dev_name;
4641 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4643 && !shash_find(ports, vlan_dev_name)) {
4644 shash_add(ports, vlan_dev_name,
4645 synthesize_splinter_port(
4646 real_dev_name, vlan_dev_name, vid));
4655 mirror_refresh_stats(struct mirror *m)
4657 struct ofproto *ofproto = m->bridge->ofproto;
4658 uint64_t tx_packets, tx_bytes;
4661 size_t stat_cnt = 0;
4663 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4664 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4668 if (tx_packets != UINT64_MAX) {
4669 keys[stat_cnt] = "tx_packets";
4670 values[stat_cnt] = tx_packets;
4673 if (tx_bytes != UINT64_MAX) {
4674 keys[stat_cnt] = "tx_bytes";
4675 values[stat_cnt] = tx_bytes;
4679 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);