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"
41 #include "ofp-print.h"
44 #include "ofproto/bond.h"
45 #include "ofproto/ofproto.h"
47 #include "poll-loop.h"
52 #include "socket-util.h"
54 #include "stream-ssl.h"
56 #include "system-stats.h"
61 #include "lib/vswitch-idl.h"
62 #include "xenserver.h"
64 #include "sflow_api.h"
65 #include "vlan-bitmap.h"
68 VLOG_DEFINE_THIS_MODULE(bridge);
70 COVERAGE_DEFINE(bridge_reconfigure);
73 /* These members are always valid.
75 * They are immutable: they never change between iface_create() and
77 struct list port_elem; /* Element in struct port's "ifaces" list. */
78 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
79 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
80 struct port *port; /* Containing port. */
81 char *name; /* Host network device name. */
82 struct netdev *netdev; /* Network device. */
83 ofp_port_t ofp_port; /* OpenFlow port number. */
86 /* These members are valid only within bridge_reconfigure(). */
87 const char *type; /* Usually same as cfg->type. */
88 const struct ovsrec_interface *cfg;
92 struct uuid uuid; /* UUID of this "mirror" record in database. */
93 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
94 struct bridge *bridge;
96 const struct ovsrec_mirror *cfg;
100 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
101 struct bridge *bridge;
104 const struct ovsrec_port *cfg;
106 /* An ordinary bridge port has 1 interface.
107 * A bridge port for bonding has at least 2 interfaces. */
108 struct list ifaces; /* List of "struct iface"s. */
112 struct hmap_node node; /* In 'all_bridges'. */
113 char *name; /* User-specified arbitrary name. */
114 char *type; /* Datapath type. */
115 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
116 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
117 const struct ovsrec_bridge *cfg;
119 /* OpenFlow switch processing. */
120 struct ofproto *ofproto; /* OpenFlow switch. */
123 struct hmap ports; /* "struct port"s indexed by name. */
124 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
125 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
127 /* Port mirroring. */
128 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
130 /* Used during reconfiguration. */
131 struct shash wanted_ports;
133 /* Synthetic local port if necessary. */
134 struct ovsrec_port synth_local_port;
135 struct ovsrec_interface synth_local_iface;
136 struct ovsrec_interface *synth_local_ifacep;
139 /* All bridges, indexed by name. */
140 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
142 /* OVSDB IDL used to obtain configuration. */
143 static struct ovsdb_idl *idl;
145 /* We want to complete daemonization, fully detaching from our parent process,
146 * only after we have completed our initial configuration, committed our state
147 * to the database, and received confirmation back from the database server
148 * that it applied the commit. This allows our parent process to know that,
149 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
150 * to have already been filled in. (It is only "very likely" rather than
151 * certain because there is always a slim possibility that the transaction will
152 * fail or that some other client has added new bridges, ports, etc. while
153 * ovs-vswitchd was configuring using an old configuration.)
155 * We only need to do this once for our initial configuration at startup, so
156 * 'initial_config_done' tracks whether we've already done it. While we are
157 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
158 * itself and is otherwise NULL. */
159 static bool initial_config_done;
160 static struct ovsdb_idl_txn *daemonize_txn;
162 /* Most recently processed IDL sequence number. */
163 static unsigned int idl_seqno;
165 /* Track changes to port connectivity. */
166 static uint64_t connectivity_seqno = LLONG_MIN;
168 /* Status update to database.
170 * Some information in the database must be kept as up-to-date as possible to
171 * allow controllers to respond rapidly to network outages. Those status are
172 * updated via the 'status_txn'.
174 * We use the global connectivity sequence number to detect the status change.
175 * Also, to prevent the status update from sending too much to the database,
176 * we check the return status of each update transaction and do not start new
177 * update if the previous transaction status is 'TXN_INCOMPLETE'.
179 * 'statux_txn' is NULL if there is no ongoing status update.
181 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
182 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
183 * which will cause the main thread wake up soon and retry the status update.
185 static struct ovsdb_idl_txn *status_txn;
186 static bool status_txn_try_again;
188 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
189 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
190 #define STATUS_CHECK_AGAIN_MSEC 100
192 /* Each time this timer expires, the bridge fetches interface and mirror
193 * statistics and pushes them into the database. */
194 static int stats_timer_interval;
195 static long long int stats_timer = LLONG_MIN;
197 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
198 * expensive. This can cause bridge_reconfigure() to take a long time during
199 * which no other work can be done. To deal with this problem, we limit port
200 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
201 * call to bridge_reconfigure(). If there is more work to do after the limit
202 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
203 * This allows the rest of the code to catch up on important things like
204 * forwarding packets. */
205 #define OFP_PORT_ACTION_WINDOW 10
207 static void add_del_bridges(const struct ovsrec_open_vswitch *);
208 static void bridge_run__(void);
209 static void bridge_create(const struct ovsrec_bridge *);
210 static void bridge_destroy(struct bridge *);
211 static struct bridge *bridge_lookup(const char *name);
212 static unixctl_cb_func bridge_unixctl_dump_flows;
213 static unixctl_cb_func bridge_unixctl_reconnect;
214 static size_t bridge_get_controllers(const struct bridge *br,
215 struct ovsrec_controller ***controllersp);
216 static void bridge_collect_wanted_ports(struct bridge *,
217 const unsigned long *splinter_vlans,
218 struct shash *wanted_ports);
219 static void bridge_delete_ofprotos(void);
220 static void bridge_delete_or_reconfigure_ports(struct bridge *);
221 static void bridge_del_ports(struct bridge *,
222 const struct shash *wanted_ports);
223 static void bridge_add_ports(struct bridge *,
224 const struct shash *wanted_ports);
226 static void bridge_configure_datapath_id(struct bridge *);
227 static void bridge_configure_netflow(struct bridge *);
228 static void bridge_configure_forward_bpdu(struct bridge *);
229 static void bridge_configure_mac_table(struct bridge *);
230 static void bridge_configure_mcast_snooping(struct bridge *);
231 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
232 static void bridge_configure_ipfix(struct bridge *);
233 static void bridge_configure_stp(struct bridge *);
234 static void bridge_configure_rstp(struct bridge *);
235 static void bridge_configure_tables(struct bridge *);
236 static void bridge_configure_dp_desc(struct bridge *);
237 static void bridge_configure_remotes(struct bridge *,
238 const struct sockaddr_in *managers,
240 static void bridge_pick_local_hw_addr(struct bridge *,
241 uint8_t ea[ETH_ADDR_LEN],
242 struct iface **hw_addr_iface);
243 static uint64_t bridge_pick_datapath_id(struct bridge *,
244 const uint8_t bridge_ea[ETH_ADDR_LEN],
245 struct iface *hw_addr_iface);
246 static uint64_t dpid_from_hash(const void *, size_t nbytes);
247 static bool bridge_has_bond_fake_iface(const struct bridge *,
249 static bool port_is_bond_fake_iface(const struct port *);
251 static unixctl_cb_func qos_unixctl_show;
253 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
254 static void port_del_ifaces(struct port *);
255 static void port_destroy(struct port *);
256 static struct port *port_lookup(const struct bridge *, const char *name);
257 static void port_configure(struct port *);
258 static struct lacp_settings *port_configure_lacp(struct port *,
259 struct lacp_settings *);
260 static void port_configure_bond(struct port *, struct bond_settings *);
261 static bool port_is_synthetic(const struct port *);
263 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
264 static void run_system_stats(void);
266 static void bridge_configure_mirrors(struct bridge *);
267 static struct mirror *mirror_create(struct bridge *,
268 const struct ovsrec_mirror *);
269 static void mirror_destroy(struct mirror *);
270 static bool mirror_configure(struct mirror *);
271 static void mirror_refresh_stats(struct mirror *);
273 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
274 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
275 const struct ovsrec_port *);
276 static bool iface_is_internal(const struct ovsrec_interface *iface,
277 const struct ovsrec_bridge *br);
278 static const char *iface_get_type(const struct ovsrec_interface *,
279 const struct ovsrec_bridge *);
280 static void iface_destroy(struct iface *);
281 static void iface_destroy__(struct iface *);
282 static struct iface *iface_lookup(const struct bridge *, const char *name);
283 static struct iface *iface_find(const char *name);
284 static struct iface *iface_from_ofp_port(const struct bridge *,
285 ofp_port_t ofp_port);
286 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
287 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
288 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
289 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
290 static void iface_configure_cfm(struct iface *);
291 static void iface_refresh_cfm_stats(struct iface *);
292 static void iface_refresh_stats(struct iface *);
293 static void iface_refresh_netdev_status(struct iface *);
294 static void iface_refresh_ofproto_status(struct iface *);
295 static bool iface_is_synthetic(const struct iface *);
296 static ofp_port_t iface_get_requested_ofp_port(
297 const struct ovsrec_interface *);
298 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
300 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
302 * This is deprecated. It is only for compatibility with broken device drivers
303 * in old versions of Linux that do not properly support VLANs when VLAN
304 * devices are not used. When broken device drivers are no longer in
305 * widespread use, we will delete these interfaces. */
307 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
308 static bool vlan_splinters_enabled_anywhere;
310 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
311 static unsigned long int *collect_splinter_vlans(
312 const struct ovsrec_open_vswitch *);
313 static void configure_splinter_port(struct port *);
314 static void add_vlan_splinter_ports(struct bridge *,
315 const unsigned long int *splinter_vlans,
316 struct shash *ports);
319 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
321 struct shash iface_hints;
322 static bool initialized = false;
329 shash_init(&iface_hints);
332 for (i = 0; i < cfg->n_bridges; i++) {
333 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
336 for (j = 0; j < br_cfg->n_ports; j++) {
337 struct ovsrec_port *port_cfg = br_cfg->ports[j];
340 for (k = 0; k < port_cfg->n_interfaces; k++) {
341 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
342 struct iface_hint *iface_hint;
344 iface_hint = xmalloc(sizeof *iface_hint);
345 iface_hint->br_name = br_cfg->name;
346 iface_hint->br_type = br_cfg->datapath_type;
347 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
349 shash_add(&iface_hints, if_cfg->name, iface_hint);
355 ofproto_init(&iface_hints);
357 shash_destroy_free_data(&iface_hints);
361 /* Public functions. */
363 /* Initializes the bridge module, configuring it to obtain its configuration
364 * from an OVSDB server accessed over 'remote', which should be a string in a
365 * form acceptable to ovsdb_idl_create(). */
367 bridge_init(const char *remote)
369 /* Create connection to database. */
370 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
371 idl_seqno = ovsdb_idl_get_seqno(idl);
372 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
373 ovsdb_idl_verify_write_only(idl);
375 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
376 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
378 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
379 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
380 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
381 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
383 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
384 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
385 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
386 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
387 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
388 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
390 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
391 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
392 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
393 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
394 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
395 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
407 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
408 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
409 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
410 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
411 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
412 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
413 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
414 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
415 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
416 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
417 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
419 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
420 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
421 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
422 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
426 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
428 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
429 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
431 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
432 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
433 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
434 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
436 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
437 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
438 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
439 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
440 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
442 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
444 /* Register unixctl commands. */
445 unixctl_command_register("qos/show", "interface", 1, 1,
446 qos_unixctl_show, NULL);
447 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
448 bridge_unixctl_dump_flows, NULL);
449 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
450 bridge_unixctl_reconnect, NULL);
462 struct bridge *br, *next_br;
464 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
467 ovsdb_idl_destroy(idl);
470 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
471 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
472 * responsible for freeing '*managersp' (with free()).
474 * You may be asking yourself "why does ovs-vswitchd care?", because
475 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
476 * should not be and in fact is not directly involved in that. But
477 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
478 * it has to tell in-band control where the managers are to enable that.
479 * (Thus, only managers connected in-band are collected.)
482 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
483 struct sockaddr_in **managersp, size_t *n_managersp)
485 struct sockaddr_in *managers = NULL;
486 size_t n_managers = 0;
490 /* Collect all of the potential targets from the "targets" columns of the
491 * rows pointed to by "manager_options", excluding any that are
494 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
495 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
497 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
498 sset_find_and_delete(&targets, m->target);
500 sset_add(&targets, m->target);
504 /* Now extract the targets' IP addresses. */
505 if (!sset_is_empty(&targets)) {
508 managers = xmalloc(sset_count(&targets) * sizeof *managers);
509 SSET_FOR_EACH (target, &targets) {
511 struct sockaddr_storage ss;
512 struct sockaddr_in in;
515 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
517 && sa.ss.ss_family == AF_INET) {
518 managers[n_managers++] = sa.in;
522 sset_destroy(&targets);
524 *managersp = managers;
525 *n_managersp = n_managers;
529 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
531 unsigned long int *splinter_vlans;
532 struct sockaddr_in *managers;
533 struct bridge *br, *next;
534 int sflow_bridge_number;
537 COVERAGE_INC(bridge_reconfigure);
539 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
540 OFPROTO_FLOW_LIMIT_DEFAULT));
541 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
542 OFPROTO_MAX_IDLE_DEFAULT));
543 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
545 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
548 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
549 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
551 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
552 * to 'ovs_cfg', with only very minimal configuration otherwise.
554 * This is mostly an update to bridge data structures. Nothing is pushed
555 * down to ofproto or lower layers. */
556 add_del_bridges(ovs_cfg);
557 splinter_vlans = collect_splinter_vlans(ovs_cfg);
558 HMAP_FOR_EACH (br, node, &all_bridges) {
559 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
560 bridge_del_ports(br, &br->wanted_ports);
562 free(splinter_vlans);
564 /* Start pushing configuration changes down to the ofproto layer:
566 * - Delete ofprotos that are no longer configured.
568 * - Delete ports that are no longer configured.
570 * - Reconfigure existing ports to their desired configurations, or
571 * delete them if not possible.
573 * We have to do all the deletions before we can do any additions, because
574 * the ports to be added might require resources that will be freed up by
575 * deletions (they might especially overlap in name). */
576 bridge_delete_ofprotos();
577 HMAP_FOR_EACH (br, node, &all_bridges) {
579 bridge_delete_or_reconfigure_ports(br);
583 /* Finish pushing configuration changes to the ofproto layer:
585 * - Create ofprotos that are missing.
587 * - Add ports that are missing. */
588 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
592 error = ofproto_create(br->name, br->type, &br->ofproto);
594 VLOG_ERR("failed to create bridge %s: %s", br->name,
595 ovs_strerror(error));
596 shash_destroy(&br->wanted_ports);
601 HMAP_FOR_EACH (br, node, &all_bridges) {
602 bridge_add_ports(br, &br->wanted_ports);
603 shash_destroy(&br->wanted_ports);
606 reconfigure_system_stats(ovs_cfg);
608 /* Complete the configuration. */
609 sflow_bridge_number = 0;
610 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
611 HMAP_FOR_EACH (br, node, &all_bridges) {
614 /* We need the datapath ID early to allow LACP ports to use it as the
615 * default system ID. */
616 bridge_configure_datapath_id(br);
618 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
621 port_configure(port);
623 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
624 iface_set_ofport(iface->cfg, iface->ofp_port);
625 /* Clear eventual previous errors */
626 ovsrec_interface_set_error(iface->cfg, NULL);
627 iface_configure_cfm(iface);
628 iface_configure_qos(iface, port->cfg->qos);
629 iface_set_mac(br, port, iface);
630 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
634 bridge_configure_mirrors(br);
635 bridge_configure_forward_bpdu(br);
636 bridge_configure_mac_table(br);
637 bridge_configure_mcast_snooping(br);
638 bridge_configure_remotes(br, managers, n_managers);
639 bridge_configure_netflow(br);
640 bridge_configure_sflow(br, &sflow_bridge_number);
641 bridge_configure_ipfix(br);
642 bridge_configure_stp(br);
643 bridge_configure_rstp(br);
644 bridge_configure_tables(br);
645 bridge_configure_dp_desc(br);
649 /* The ofproto-dpif provider does some final reconfiguration in its
650 * ->type_run() function. We have to call it before notifying the database
651 * client that reconfiguration is complete, otherwise there is a very
652 * narrow race window in which e.g. ofproto/trace will not recognize the
653 * new configuration (sometimes this causes unit test failures). */
657 /* Delete ofprotos which aren't configured or have the wrong type. Create
658 * ofprotos which don't exist but need to. */
660 bridge_delete_ofprotos(void)
667 /* Delete ofprotos with no bridge or with the wrong type. */
670 ofproto_enumerate_types(&types);
671 SSET_FOR_EACH (type, &types) {
674 ofproto_enumerate_names(type, &names);
675 SSET_FOR_EACH (name, &names) {
676 br = bridge_lookup(name);
677 if (!br || strcmp(type, br->type)) {
678 ofproto_delete(name, type);
682 sset_destroy(&names);
683 sset_destroy(&types);
687 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
689 if (*n >= *allocated) {
690 ports = x2nrealloc(ports, allocated, sizeof *ports);
692 ports[(*n)++] = port;
697 bridge_delete_or_reconfigure_ports(struct bridge *br)
699 struct ofproto_port ofproto_port;
700 struct ofproto_port_dump dump;
702 struct sset ofproto_ports;
703 struct port *port, *port_next;
705 /* List of "ofp_port"s to delete. We make a list instead of deleting them
706 * right away because ofproto implementations aren't necessarily able to
707 * iterate through a changing list of ports in an entirely robust way. */
714 sset_init(&ofproto_ports);
716 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
717 * that are not configured in the database. (This commonly happens when
718 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
720 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
721 * that have the wrong OpenFlow port number (and arrange to add them back
722 * with the correct OpenFlow port number). */
723 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
724 ofp_port_t requested_ofp_port;
727 sset_add(&ofproto_ports, ofproto_port.name);
729 iface = iface_lookup(br, ofproto_port.name);
731 /* No such iface is configured, so we should delete this
734 * As a corner case exception, keep the port if it's a bond fake
736 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
737 && !strcmp(ofproto_port.type, "internal")) {
743 if (strcmp(ofproto_port.type, iface->type)
744 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
745 /* The interface is the wrong type or can't be configured.
750 /* If the requested OpenFlow port for 'iface' changed, and it's not
751 * already the correct port, then we might want to temporarily delete
752 * this interface, so we can add it back again with the new OpenFlow
754 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
755 if (iface->ofp_port != OFPP_LOCAL &&
756 requested_ofp_port != OFPP_NONE &&
757 requested_ofp_port != iface->ofp_port) {
758 ofp_port_t victim_request;
759 struct iface *victim;
761 /* Check for an existing OpenFlow port currently occupying
762 * 'iface''s requested port number. If there isn't one, then
763 * delete this port. Otherwise we need to consider further. */
764 victim = iface_from_ofp_port(br, requested_ofp_port);
769 /* 'victim' is a port currently using 'iface''s requested port
770 * number. Unless 'victim' specifically requested that port
771 * number, too, then we can delete both 'iface' and 'victim'
772 * temporarily. (We'll add both of them back again later with new
773 * OpenFlow port numbers.)
775 * If 'victim' did request port number 'requested_ofp_port', just
776 * like 'iface', then that's a configuration inconsistency that we
777 * can't resolve. We might as well let it keep its current port
779 victim_request = iface_get_requested_ofp_port(victim->cfg);
780 if (victim_request != requested_ofp_port) {
781 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
782 iface_destroy(victim);
791 iface_destroy(iface);
792 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
794 for (i = 0; i < n; i++) {
795 ofproto_port_del(br->ofproto, del[i]);
799 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
800 * that we didn't see when we iterated through the datapath, i.e. ports
801 * that disappeared underneath use. This is an unusual situation, but it
802 * can happen in some cases:
804 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
805 * idea but could happen).
807 * - The port represented a device that disappeared, e.g. a tuntap
808 * device destroyed via "tunctl -d", a physical Ethernet device
809 * whose module was just unloaded via "rmmod", or a virtual NIC for a
810 * VM whose VM was just terminated. */
811 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
812 struct iface *iface, *iface_next;
814 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
815 if (!sset_contains(&ofproto_ports, iface->name)) {
816 iface_destroy__(iface);
820 if (list_is_empty(&port->ifaces)) {
824 sset_destroy(&ofproto_ports);
828 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
829 bool with_requested_port)
831 struct shash_node *port_node;
833 SHASH_FOR_EACH (port_node, wanted_ports) {
834 const struct ovsrec_port *port_cfg = port_node->data;
837 for (i = 0; i < port_cfg->n_interfaces; i++) {
838 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
839 ofp_port_t requested_ofp_port;
841 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
842 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
843 struct iface *iface = iface_lookup(br, iface_cfg->name);
846 iface_create(br, iface_cfg, port_cfg);
854 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
856 /* First add interfaces that request a particular port number. */
857 bridge_add_ports__(br, wanted_ports, true);
859 /* Then add interfaces that want automatic port number assignment.
860 * We add these afterward to avoid accidentally taking a specifically
861 * requested port number. */
862 bridge_add_ports__(br, wanted_ports, false);
866 port_configure(struct port *port)
868 const struct ovsrec_port *cfg = port->cfg;
869 struct bond_settings bond_settings;
870 struct lacp_settings lacp_settings;
871 struct ofproto_bundle_settings s;
874 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
875 configure_splinter_port(port);
884 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
885 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
886 s.slaves[s.n_slaves++] = iface->ofp_port;
891 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
895 /* Get VLAN trunks. */
898 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
902 if (cfg->vlan_mode) {
903 if (!strcmp(cfg->vlan_mode, "access")) {
904 s.vlan_mode = PORT_VLAN_ACCESS;
905 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
906 s.vlan_mode = PORT_VLAN_TRUNK;
907 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
908 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
909 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
910 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
912 /* This "can't happen" because ovsdb-server should prevent it. */
913 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
914 "back to trunk mode", port->name, cfg->vlan_mode);
915 s.vlan_mode = PORT_VLAN_TRUNK;
919 s.vlan_mode = PORT_VLAN_ACCESS;
921 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
925 s.vlan_mode = PORT_VLAN_TRUNK;
928 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
931 /* Get LACP settings. */
932 s.lacp = port_configure_lacp(port, &lacp_settings);
936 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
937 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
938 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
941 s.lacp_slaves = NULL;
944 /* Get bond settings. */
945 if (s.n_slaves > 1) {
946 s.bond = &bond_settings;
947 port_configure_bond(port, &bond_settings);
950 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
951 netdev_set_miimon_interval(iface->netdev, 0);
956 ofproto_bundle_register(port->bridge->ofproto, port, &s);
964 /* Pick local port hardware address and datapath ID for 'br'. */
966 bridge_configure_datapath_id(struct bridge *br)
968 uint8_t ea[ETH_ADDR_LEN];
970 struct iface *local_iface;
971 struct iface *hw_addr_iface;
974 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
975 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
977 int error = netdev_set_etheraddr(local_iface->netdev, ea);
979 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
980 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
981 "Ethernet address: %s",
982 br->name, ovs_strerror(error));
985 memcpy(br->ea, ea, ETH_ADDR_LEN);
987 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
988 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
989 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
990 ofproto_set_datapath_id(br->ofproto, dpid);
993 dpid_string = xasprintf("%016"PRIx64, dpid);
994 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
998 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1001 bridge_get_allowed_versions(struct bridge *br)
1003 if (!br->cfg->n_protocols)
1006 return ofputil_versions_from_strings(br->cfg->protocols,
1007 br->cfg->n_protocols);
1010 /* Set NetFlow configuration on 'br'. */
1012 bridge_configure_netflow(struct bridge *br)
1014 struct ovsrec_netflow *cfg = br->cfg->netflow;
1015 struct netflow_options opts;
1018 ofproto_set_netflow(br->ofproto, NULL);
1022 memset(&opts, 0, sizeof opts);
1024 /* Get default NetFlow configuration from datapath.
1025 * Apply overrides from 'cfg'. */
1026 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1027 if (cfg->engine_type) {
1028 opts.engine_type = *cfg->engine_type;
1030 if (cfg->engine_id) {
1031 opts.engine_id = *cfg->engine_id;
1034 /* Configure active timeout interval. */
1035 opts.active_timeout = cfg->active_timeout;
1036 if (!opts.active_timeout) {
1037 opts.active_timeout = -1;
1038 } else if (opts.active_timeout < 0) {
1039 VLOG_WARN("bridge %s: active timeout interval set to negative "
1040 "value, using default instead (%d seconds)", br->name,
1041 NF_ACTIVE_TIMEOUT_DEFAULT);
1042 opts.active_timeout = -1;
1045 /* Add engine ID to interface number to disambiguate bridgs? */
1046 opts.add_id_to_iface = cfg->add_id_to_interface;
1047 if (opts.add_id_to_iface) {
1048 if (opts.engine_id > 0x7f) {
1049 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1050 "another vswitch, choose an engine id less than 128",
1053 if (hmap_count(&br->ports) > 508) {
1054 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1055 "another port when more than 508 ports are used",
1061 sset_init(&opts.collectors);
1062 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1065 if (ofproto_set_netflow(br->ofproto, &opts)) {
1066 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1068 sset_destroy(&opts.collectors);
1071 /* Set sFlow configuration on 'br'. */
1073 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1075 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1076 struct ovsrec_controller **controllers;
1077 struct ofproto_sflow_options oso;
1078 size_t n_controllers;
1082 ofproto_set_sflow(br->ofproto, NULL);
1086 memset(&oso, 0, sizeof oso);
1088 sset_init(&oso.targets);
1089 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1091 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1092 if (cfg->sampling) {
1093 oso.sampling_rate = *cfg->sampling;
1096 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1098 oso.polling_interval = *cfg->polling;
1101 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1103 oso.header_len = *cfg->header;
1106 oso.sub_id = (*sflow_bridge_number)++;
1107 oso.agent_device = cfg->agent;
1109 oso.control_ip = NULL;
1110 n_controllers = bridge_get_controllers(br, &controllers);
1111 for (i = 0; i < n_controllers; i++) {
1112 if (controllers[i]->local_ip) {
1113 oso.control_ip = controllers[i]->local_ip;
1117 ofproto_set_sflow(br->ofproto, &oso);
1119 sset_destroy(&oso.targets);
1122 /* Returns whether a IPFIX row is valid. */
1124 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1126 return ipfix && ipfix->n_targets > 0;
1129 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1131 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1132 const struct bridge *br)
1134 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1137 /* Set IPFIX configuration on 'br'. */
1139 bridge_configure_ipfix(struct bridge *br)
1141 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1142 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1143 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1144 struct ofproto_ipfix_bridge_exporter_options be_opts;
1145 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1146 size_t n_fe_opts = 0;
1148 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1149 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1154 if (!valid_be_cfg && n_fe_opts == 0) {
1155 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1160 memset(&be_opts, 0, sizeof be_opts);
1162 sset_init(&be_opts.targets);
1163 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1165 if (be_cfg->sampling) {
1166 be_opts.sampling_rate = *be_cfg->sampling;
1168 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1170 if (be_cfg->obs_domain_id) {
1171 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1173 if (be_cfg->obs_point_id) {
1174 be_opts.obs_point_id = *be_cfg->obs_point_id;
1176 if (be_cfg->cache_active_timeout) {
1177 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1179 if (be_cfg->cache_max_flows) {
1180 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1183 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1184 "enable-tunnel-sampling", true);
1186 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1187 "enable-input-sampling", false);
1189 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1190 "enable-output-sampling", false);
1193 if (n_fe_opts > 0) {
1194 struct ofproto_ipfix_flow_exporter_options *opts;
1195 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1197 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1198 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1199 opts->collector_set_id = fe_cfg->id;
1200 sset_init(&opts->targets);
1201 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1202 fe_cfg->ipfix->n_targets);
1203 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1204 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1205 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1206 ? *fe_cfg->ipfix->cache_max_flows : 0;
1212 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1216 sset_destroy(&be_opts.targets);
1219 if (n_fe_opts > 0) {
1220 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1222 for (i = 0; i < n_fe_opts; i++) {
1223 sset_destroy(&opts->targets);
1231 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1232 struct ofproto_port_stp_settings *port_s,
1233 int *port_num_counter, unsigned long *port_num_bitmap)
1235 const char *config_str;
1236 struct iface *iface;
1238 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1239 port_s->enable = false;
1242 port_s->enable = true;
1245 /* STP over bonds is not supported. */
1246 if (!list_is_singleton(&port->ifaces)) {
1247 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1249 port_s->enable = false;
1253 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1255 /* Internal ports shouldn't participate in spanning tree, so
1257 if (!strcmp(iface->type, "internal")) {
1258 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1259 port_s->enable = false;
1263 /* STP on mirror output ports is not supported. */
1264 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1265 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1266 port_s->enable = false;
1270 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1272 unsigned long int port_num = strtoul(config_str, NULL, 0);
1273 int port_idx = port_num - 1;
1275 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1276 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1277 port_s->enable = false;
1281 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1282 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1283 port->name, port_num);
1284 port_s->enable = false;
1287 bitmap_set1(port_num_bitmap, port_idx);
1288 port_s->port_num = port_idx;
1290 if (*port_num_counter >= STP_MAX_PORTS) {
1291 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1292 port_s->enable = false;
1296 port_s->port_num = (*port_num_counter)++;
1299 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1301 port_s->path_cost = strtoul(config_str, NULL, 10);
1303 enum netdev_features current;
1306 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1307 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1308 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1311 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1313 port_s->priority = strtoul(config_str, NULL, 0);
1315 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1320 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1321 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1323 const char *config_str;
1324 struct iface *iface;
1326 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1327 port_s->enable = false;
1330 port_s->enable = true;
1333 /* RSTP over bonds is not supported. */
1334 if (!list_is_singleton(&port->ifaces)) {
1335 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1337 port_s->enable = false;
1341 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1343 /* Internal ports shouldn't participate in spanning tree, so
1345 if (!strcmp(iface->type, "internal")) {
1346 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1347 port_s->enable = false;
1351 /* RSTP on mirror output ports is not supported. */
1352 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1353 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1354 port_s->enable = false;
1358 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1360 unsigned long int port_num = strtoul(config_str, NULL, 0);
1361 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1362 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1363 port_s->enable = false;
1366 port_s->port_num = port_num;
1368 if (*port_num_counter >= RSTP_MAX_PORTS) {
1369 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1370 port_s->enable = false;
1373 /* If rstp-port-num is not specified, use 0.
1374 * rstp_port_set_port_number() will look for the first free one. */
1375 port_s->port_num = 0;
1378 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1380 port_s->path_cost = strtoul(config_str, NULL, 10);
1382 enum netdev_features current;
1385 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1386 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1387 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1390 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1392 port_s->priority = strtoul(config_str, NULL, 0);
1394 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1397 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1398 "rstp-port-admin-edge", false);
1399 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1400 "rstp-port-auto-edge", true);
1401 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1402 "rstp-port-mcheck", false);
1405 /* Set spanning tree configuration on 'br'. */
1407 bridge_configure_stp(struct bridge *br)
1409 struct ofproto_rstp_status rstp_status;
1411 ofproto_get_rstp_status(br->ofproto, &rstp_status);
1412 if (!br->cfg->stp_enable) {
1413 ofproto_set_stp(br->ofproto, NULL);
1414 } else if (rstp_status.enabled) {
1415 /* Do not activate STP if RSTP is enabled. */
1416 VLOG_ERR("STP cannot be enabled if RSTP is running.");
1417 ofproto_set_stp(br->ofproto, NULL);
1418 ovsrec_bridge_set_stp_enable(br->cfg, false);
1420 struct ofproto_stp_settings br_s;
1421 const char *config_str;
1423 int port_num_counter;
1424 unsigned long *port_num_bitmap;
1426 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1428 uint8_t ea[ETH_ADDR_LEN];
1430 if (eth_addr_from_string(config_str, ea)) {
1431 br_s.system_id = eth_addr_to_uint64(ea);
1433 br_s.system_id = eth_addr_to_uint64(br->ea);
1434 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1435 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1438 br_s.system_id = eth_addr_to_uint64(br->ea);
1441 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1443 br_s.priority = strtoul(config_str, NULL, 0);
1445 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1448 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1450 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1452 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1455 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1457 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1459 br_s.max_age = STP_DEFAULT_MAX_AGE;
1462 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1464 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1466 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1469 /* Configure STP on the bridge. */
1470 if (ofproto_set_stp(br->ofproto, &br_s)) {
1471 VLOG_ERR("bridge %s: could not enable STP", br->name);
1475 /* Users must either set the port number with the "stp-port-num"
1476 * configuration on all ports or none. If manual configuration
1477 * is not done, then we allocate them sequentially. */
1478 port_num_counter = 0;
1479 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1480 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1481 struct ofproto_port_stp_settings port_s;
1482 struct iface *iface;
1484 port_configure_stp(br->ofproto, port, &port_s,
1485 &port_num_counter, port_num_bitmap);
1487 /* As bonds are not supported, just apply configuration to
1488 * all interfaces. */
1489 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1490 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1492 VLOG_ERR("port %s: could not enable STP", port->name);
1498 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1499 && port_num_counter) {
1500 VLOG_ERR("bridge %s: must manually configure all STP port "
1501 "IDs or none, disabling", br->name);
1502 ofproto_set_stp(br->ofproto, NULL);
1504 bitmap_free(port_num_bitmap);
1509 bridge_configure_rstp(struct bridge *br)
1511 struct ofproto_stp_status stp_status;
1513 ofproto_get_stp_status(br->ofproto, &stp_status);
1514 if (!br->cfg->rstp_enable) {
1515 ofproto_set_rstp(br->ofproto, NULL);
1516 } else if (stp_status.enabled) {
1517 /* Do not activate RSTP if STP is enabled. */
1518 VLOG_ERR("RSTP cannot be enabled if STP is running.");
1519 ofproto_set_rstp(br->ofproto, NULL);
1520 ovsrec_bridge_set_rstp_enable(br->cfg, false);
1522 struct ofproto_rstp_settings br_s;
1523 const char *config_str;
1525 int port_num_counter;
1527 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1529 uint8_t ea[ETH_ADDR_LEN];
1531 if (eth_addr_from_string(config_str, ea)) {
1532 br_s.address = eth_addr_to_uint64(ea);
1535 br_s.address = eth_addr_to_uint64(br->ea);
1536 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1537 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1541 br_s.address = eth_addr_to_uint64(br->ea);
1544 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1546 br_s.priority = strtoul(config_str, NULL, 0);
1548 br_s.priority = RSTP_DEFAULT_PRIORITY;
1551 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1553 br_s.ageing_time = strtoul(config_str, NULL, 0);
1555 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1558 config_str = smap_get(&br->cfg->other_config,
1559 "rstp-force-protocol-version");
1561 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1563 br_s.force_protocol_version = FPV_DEFAULT;
1566 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1568 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1570 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1573 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1575 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1577 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1580 config_str = smap_get(&br->cfg->other_config,
1581 "rstp-transmit-hold-count");
1583 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1585 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1588 /* Configure RSTP on the bridge. */
1589 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1590 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1594 port_num_counter = 0;
1595 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1596 struct ofproto_port_rstp_settings port_s;
1597 struct iface *iface;
1599 port_configure_rstp(br->ofproto, port, &port_s,
1602 /* As bonds are not supported, just apply configuration to
1603 * all interfaces. */
1604 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1605 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1607 VLOG_ERR("port %s: could not enable RSTP", port->name);
1616 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1618 const struct port *port = port_lookup(br, name);
1619 return port && port_is_bond_fake_iface(port);
1623 port_is_bond_fake_iface(const struct port *port)
1625 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1629 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1631 struct bridge *br, *next;
1632 struct shash new_br;
1635 /* Collect new bridges' names and types. */
1636 shash_init(&new_br);
1637 for (i = 0; i < cfg->n_bridges; i++) {
1638 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1639 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1641 if (strchr(br_cfg->name, '/')) {
1642 /* Prevent remote ovsdb-server users from accessing arbitrary
1643 * directories, e.g. consider a bridge named "../../../etc/". */
1644 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1646 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1647 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1651 /* Get rid of deleted bridges or those whose types have changed.
1652 * Update 'cfg' of bridges that still exist. */
1653 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1654 br->cfg = shash_find_data(&new_br, br->name);
1655 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1656 br->cfg->datapath_type))) {
1661 /* Add new bridges. */
1662 for (i = 0; i < cfg->n_bridges; i++) {
1663 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1664 struct bridge *br = bridge_lookup(br_cfg->name);
1666 bridge_create(br_cfg);
1670 shash_destroy(&new_br);
1673 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1674 * Returns 0 if successful, otherwise a positive errno value. */
1676 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1677 struct netdev *netdev, char **errp)
1679 return netdev_set_config(netdev, &iface_cfg->options, errp);
1682 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1683 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1685 * If successful, returns 0 and stores the network device in '*netdevp'. On
1686 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1688 iface_do_create(const struct bridge *br,
1689 const struct ovsrec_interface *iface_cfg,
1690 const struct ovsrec_port *port_cfg,
1691 ofp_port_t *ofp_portp, struct netdev **netdevp,
1694 struct netdev *netdev = NULL;
1697 if (netdev_is_reserved_name(iface_cfg->name)) {
1698 VLOG_WARN("could not create interface %s, name is reserved",
1704 error = netdev_open(iface_cfg->name,
1705 iface_get_type(iface_cfg, br->cfg), &netdev);
1707 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1708 iface_cfg->name, ovs_strerror(error));
1712 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1717 *ofp_portp = iface_pick_ofport(iface_cfg);
1718 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1723 VLOG_INFO("bridge %s: added interface %s on port %d",
1724 br->name, iface_cfg->name, *ofp_portp);
1726 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1727 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1735 netdev_close(netdev);
1739 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1740 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1741 * automatically allocated for the iface. Takes ownership of and
1742 * deallocates 'if_cfg'.
1744 * Return true if an iface is successfully created, false otherwise. */
1746 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1747 const struct ovsrec_port *port_cfg)
1749 struct netdev *netdev;
1750 struct iface *iface;
1751 ofp_port_t ofp_port;
1756 /* Do the bits that can fail up front. */
1757 ovs_assert(!iface_lookup(br, iface_cfg->name));
1758 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1760 iface_clear_db_record(iface_cfg, errp);
1765 /* Get or create the port structure. */
1766 port = port_lookup(br, port_cfg->name);
1768 port = port_create(br, port_cfg);
1771 /* Create the iface structure. */
1772 iface = xzalloc(sizeof *iface);
1773 list_push_back(&port->ifaces, &iface->port_elem);
1774 hmap_insert(&br->iface_by_name, &iface->name_node,
1775 hash_string(iface_cfg->name, 0));
1777 iface->name = xstrdup(iface_cfg->name);
1778 iface->ofp_port = ofp_port;
1779 iface->netdev = netdev;
1780 iface->type = iface_get_type(iface_cfg, br->cfg);
1781 iface->cfg = iface_cfg;
1782 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1783 hash_ofp_port(ofp_port));
1785 /* Populate initial status in database. */
1786 iface_refresh_stats(iface);
1787 iface_refresh_netdev_status(iface);
1789 /* Add bond fake iface if necessary. */
1790 if (port_is_bond_fake_iface(port)) {
1791 struct ofproto_port ofproto_port;
1793 if (ofproto_port_query_by_name(br->ofproto, port->name,
1795 struct netdev *netdev;
1798 error = netdev_open(port->name, "internal", &netdev);
1800 ofp_port_t fake_ofp_port = OFPP_NONE;
1801 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1802 netdev_close(netdev);
1804 VLOG_WARN("could not open network device %s (%s)",
1805 port->name, ovs_strerror(error));
1808 /* Already exists, nothing to do. */
1809 ofproto_port_destroy(&ofproto_port);
1816 /* Set forward BPDU option. */
1818 bridge_configure_forward_bpdu(struct bridge *br)
1820 ofproto_set_forward_bpdu(br->ofproto,
1821 smap_get_bool(&br->cfg->other_config,
1826 /* Set MAC learning table configuration for 'br'. */
1828 bridge_configure_mac_table(struct bridge *br)
1830 const char *idle_time_str;
1833 const char *mac_table_size_str;
1836 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1837 idle_time = (idle_time_str && atoi(idle_time_str)
1838 ? atoi(idle_time_str)
1839 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1841 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1842 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1843 ? atoi(mac_table_size_str)
1846 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1849 /* Set multicast snooping table configuration for 'br'. */
1851 bridge_configure_mcast_snooping(struct bridge *br)
1853 if (!br->cfg->mcast_snooping_enable) {
1854 ofproto_set_mcast_snooping(br->ofproto, NULL);
1857 struct ofproto_mcast_snooping_settings br_s;
1858 const char *idle_time_str;
1859 const char *max_entries_str;
1861 idle_time_str = smap_get(&br->cfg->other_config,
1862 "mcast-snooping-aging-time");
1863 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1864 ? atoi(idle_time_str)
1865 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1867 max_entries_str = smap_get(&br->cfg->other_config,
1868 "mcast-snooping-table-size");
1869 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1870 ? atoi(max_entries_str)
1871 : MCAST_DEFAULT_MAX_ENTRIES);
1873 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1874 "mcast-snooping-disable-flood-unregistered",
1877 /* Configure multicast snooping on the bridge */
1878 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1879 VLOG_ERR("bridge %s: could not enable multicast snooping",
1884 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1885 bool flood = smap_get_bool(&port->cfg->other_config,
1886 "mcast-snooping-flood", false);
1887 if (ofproto_port_set_mcast_snooping(br->ofproto, port, flood)) {
1888 VLOG_ERR("port %s: could not configure mcast snooping",
1896 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1897 const struct port *fake_br, struct iface **hw_addr_iface)
1899 struct hmapx mirror_output_ports;
1901 bool found_addr = false;
1905 /* Mirror output ports don't participate in picking the local hardware
1906 * address. ofproto can't help us find out whether a given port is a
1907 * mirror output because we haven't configured mirrors yet, so we need to
1908 * accumulate them ourselves. */
1909 hmapx_init(&mirror_output_ports);
1910 for (i = 0; i < br->cfg->n_mirrors; i++) {
1911 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1912 if (m->output_port) {
1913 hmapx_add(&mirror_output_ports, m->output_port);
1917 /* Otherwise choose the minimum non-local MAC address among all of the
1919 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1920 uint8_t iface_ea[ETH_ADDR_LEN];
1921 struct iface *candidate;
1922 struct iface *iface;
1924 /* Mirror output ports don't participate. */
1925 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1929 /* Choose the MAC address to represent the port. */
1931 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1932 /* Find the interface with this Ethernet address (if any) so that
1933 * we can provide the correct devname to the caller. */
1934 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1935 uint8_t candidate_ea[ETH_ADDR_LEN];
1936 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1937 && eth_addr_equals(iface_ea, candidate_ea)) {
1942 /* Choose the interface whose MAC address will represent the port.
1943 * The Linux kernel bonding code always chooses the MAC address of
1944 * the first slave added to a bond, and the Fedora networking
1945 * scripts always add slaves to a bond in alphabetical order, so
1946 * for compatibility we choose the interface with the name that is
1947 * first in alphabetical order. */
1948 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1949 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1954 /* The local port doesn't count (since we're trying to choose its
1955 * MAC address anyway). */
1956 if (iface->ofp_port == OFPP_LOCAL) {
1960 /* For fake bridges we only choose from ports with the same tag */
1961 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1962 if (!port->cfg->tag) {
1965 if (*port->cfg->tag != *fake_br->cfg->tag) {
1971 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1977 /* Compare against our current choice. */
1978 if (!eth_addr_is_multicast(iface_ea) &&
1979 !eth_addr_is_local(iface_ea) &&
1980 !eth_addr_is_reserved(iface_ea) &&
1981 !eth_addr_is_zero(iface_ea) &&
1982 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1984 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1985 *hw_addr_iface = iface;
1991 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1992 *hw_addr_iface = NULL;
1995 hmapx_destroy(&mirror_output_ports);
1999 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2000 struct iface **hw_addr_iface)
2003 *hw_addr_iface = NULL;
2005 /* Did the user request a particular MAC? */
2006 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2007 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2008 if (eth_addr_is_multicast(ea)) {
2009 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2010 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2011 } else if (eth_addr_is_zero(ea)) {
2012 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2018 /* Find a local hw address */
2019 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2022 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2023 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2024 * an interface on 'br', then that interface must be passed in as
2025 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2026 * 'hw_addr_iface' must be passed in as a null pointer. */
2028 bridge_pick_datapath_id(struct bridge *br,
2029 const uint8_t bridge_ea[ETH_ADDR_LEN],
2030 struct iface *hw_addr_iface)
2033 * The procedure for choosing a bridge MAC address will, in the most
2034 * ordinary case, also choose a unique MAC that we can use as a datapath
2035 * ID. In some special cases, though, multiple bridges will end up with
2036 * the same MAC address. This is OK for the bridges, but it will confuse
2037 * the OpenFlow controller, because each datapath needs a unique datapath
2040 * Datapath IDs must be unique. It is also very desirable that they be
2041 * stable from one run to the next, so that policy set on a datapath
2044 const char *datapath_id;
2047 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2048 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2052 if (!hw_addr_iface) {
2054 * A purely internal bridge, that is, one that has no non-virtual
2055 * network devices on it at all, is difficult because it has no
2056 * natural unique identifier at all.
2058 * When the host is a XenServer, we handle this case by hashing the
2059 * host's UUID with the name of the bridge. Names of bridges are
2060 * persistent across XenServer reboots, although they can be reused if
2061 * an internal network is destroyed and then a new one is later
2062 * created, so this is fairly effective.
2064 * When the host is not a XenServer, we punt by using a random MAC
2065 * address on each run.
2067 const char *host_uuid = xenserver_get_host_uuid();
2069 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2070 dpid = dpid_from_hash(combined, strlen(combined));
2076 return eth_addr_to_uint64(bridge_ea);
2080 dpid_from_hash(const void *data, size_t n)
2082 uint8_t hash[SHA1_DIGEST_SIZE];
2084 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2085 sha1_bytes(data, n, hash);
2086 eth_addr_mark_random(hash);
2087 return eth_addr_to_uint64(hash);
2091 iface_refresh_netdev_status(struct iface *iface)
2095 enum netdev_features current;
2096 enum netdev_flags flags;
2097 const char *link_state;
2098 uint8_t mac[ETH_ADDR_LEN];
2099 int64_t bps, mtu_64, ifindex64, link_resets;
2102 if (iface_is_synthetic(iface)) {
2106 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2107 && !status_txn_try_again) {
2111 iface->change_seq = netdev_get_change_seq(iface->netdev);
2115 if (!netdev_get_status(iface->netdev, &smap)) {
2116 ovsrec_interface_set_status(iface->cfg, &smap);
2118 ovsrec_interface_set_status(iface->cfg, NULL);
2121 smap_destroy(&smap);
2123 error = netdev_get_flags(iface->netdev, &flags);
2125 const char *state = flags & NETDEV_UP ? "up" : "down";
2127 ovsrec_interface_set_admin_state(iface->cfg, state);
2129 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2132 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2133 ovsrec_interface_set_link_state(iface->cfg, link_state);
2135 link_resets = netdev_get_carrier_resets(iface->netdev);
2136 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2138 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2139 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2141 ovsrec_interface_set_duplex(iface->cfg,
2142 netdev_features_is_full_duplex(current)
2144 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2146 ovsrec_interface_set_duplex(iface->cfg, NULL);
2147 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2150 error = netdev_get_mtu(iface->netdev, &mtu);
2153 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2155 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2158 error = netdev_get_etheraddr(iface->netdev, mac);
2160 char mac_string[32];
2162 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2163 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2165 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2168 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2169 * if there is no valid ifindex number. */
2170 ifindex64 = netdev_get_ifindex(iface->netdev);
2171 if (ifindex64 < 0) {
2174 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2178 iface_refresh_ofproto_status(struct iface *iface)
2182 if (iface_is_synthetic(iface)) {
2186 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2190 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2192 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2195 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2197 || status_txn_try_again) {
2198 iface_refresh_cfm_stats(iface);
2201 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2203 || status_txn_try_again) {
2207 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2208 iface->ofp_port, &smap);
2209 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2210 smap_destroy(&smap);
2214 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2217 iface_refresh_cfm_stats(struct iface *iface)
2219 const struct ovsrec_interface *cfg = iface->cfg;
2220 struct cfm_status status;
2223 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2224 iface->ofp_port, &status);
2226 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2227 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2228 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2229 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2230 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2231 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2233 const char *reasons[CFM_FAULT_N_REASONS];
2234 int64_t cfm_health = status.health;
2235 int64_t cfm_flap_count = status.flap_count;
2236 bool faulted = status.faults != 0;
2239 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2242 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2243 int reason = 1 << i;
2244 if (status.faults & reason) {
2245 reasons[j++] = cfm_fault_reason_to_str(reason);
2248 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
2250 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2252 if (status.remote_opstate >= 0) {
2253 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2254 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2256 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2259 ovsrec_interface_set_cfm_remote_mpids(cfg,
2260 (const int64_t *)status.rmps,
2262 if (cfm_health >= 0) {
2263 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2265 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2273 iface_refresh_stats(struct iface *iface)
2275 #define IFACE_STATS \
2276 IFACE_STAT(rx_packets, "rx_packets") \
2277 IFACE_STAT(tx_packets, "tx_packets") \
2278 IFACE_STAT(rx_bytes, "rx_bytes") \
2279 IFACE_STAT(tx_bytes, "tx_bytes") \
2280 IFACE_STAT(rx_dropped, "rx_dropped") \
2281 IFACE_STAT(tx_dropped, "tx_dropped") \
2282 IFACE_STAT(rx_errors, "rx_errors") \
2283 IFACE_STAT(tx_errors, "tx_errors") \
2284 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2285 IFACE_STAT(rx_over_errors, "rx_over_err") \
2286 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2287 IFACE_STAT(collisions, "collisions")
2289 #define IFACE_STAT(MEMBER, NAME) + 1
2290 enum { N_IFACE_STATS = IFACE_STATS };
2292 int64_t values[N_IFACE_STATS];
2293 char *keys[N_IFACE_STATS];
2296 struct netdev_stats stats;
2298 if (iface_is_synthetic(iface)) {
2302 /* Intentionally ignore return value, since errors will set 'stats' to
2303 * all-1s, and we will deal with that correctly below. */
2304 netdev_get_stats(iface->netdev, &stats);
2306 /* Copy statistics into keys[] and values[]. */
2308 #define IFACE_STAT(MEMBER, NAME) \
2309 if (stats.MEMBER != UINT64_MAX) { \
2311 values[n] = stats.MEMBER; \
2316 ovs_assert(n <= N_IFACE_STATS);
2318 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2323 br_refresh_stp_status(struct bridge *br)
2325 struct smap smap = SMAP_INITIALIZER(&smap);
2326 struct ofproto *ofproto = br->ofproto;
2327 struct ofproto_stp_status status;
2329 if (ofproto_get_stp_status(ofproto, &status)) {
2333 if (!status.enabled) {
2334 ovsrec_bridge_set_status(br->cfg, NULL);
2338 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2339 STP_ID_ARGS(status.bridge_id));
2340 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2341 STP_ID_ARGS(status.designated_root));
2342 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2344 ovsrec_bridge_set_status(br->cfg, &smap);
2345 smap_destroy(&smap);
2349 port_refresh_stp_status(struct port *port)
2351 struct ofproto *ofproto = port->bridge->ofproto;
2352 struct iface *iface;
2353 struct ofproto_port_stp_status status;
2356 if (port_is_synthetic(port)) {
2360 /* STP doesn't currently support bonds. */
2361 if (!list_is_singleton(&port->ifaces)) {
2362 ovsrec_port_set_status(port->cfg, NULL);
2366 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2367 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2371 if (!status.enabled) {
2372 ovsrec_port_set_status(port->cfg, NULL);
2376 /* Set Status column. */
2378 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2379 smap_add(&smap, "stp_state", stp_state_name(status.state));
2380 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2381 smap_add(&smap, "stp_role", stp_role_name(status.role));
2382 ovsrec_port_set_status(port->cfg, &smap);
2383 smap_destroy(&smap);
2387 port_refresh_stp_stats(struct port *port)
2389 struct ofproto *ofproto = port->bridge->ofproto;
2390 struct iface *iface;
2391 struct ofproto_port_stp_stats stats;
2393 int64_t int_values[3];
2395 if (port_is_synthetic(port)) {
2399 /* STP doesn't currently support bonds. */
2400 if (!list_is_singleton(&port->ifaces)) {
2404 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2405 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2409 if (!stats.enabled) {
2410 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2414 /* Set Statistics column. */
2415 keys[0] = "stp_tx_count";
2416 int_values[0] = stats.tx_count;
2417 keys[1] = "stp_rx_count";
2418 int_values[1] = stats.rx_count;
2419 keys[2] = "stp_error_count";
2420 int_values[2] = stats.error_count;
2422 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2423 ARRAY_SIZE(int_values));
2427 br_refresh_rstp_status(struct bridge *br)
2429 struct smap smap = SMAP_INITIALIZER(&smap);
2430 struct ofproto *ofproto = br->ofproto;
2431 struct ofproto_rstp_status status;
2433 if (ofproto_get_rstp_status(ofproto, &status)) {
2436 if (!status.enabled) {
2437 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2440 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2441 RSTP_ID_ARGS(status.bridge_id));
2442 smap_add_format(&smap, "rstp_root_path_cost", "%d",
2443 status.root_path_cost);
2444 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2445 RSTP_ID_ARGS(status.root_id));
2446 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2447 RSTP_ID_ARGS(status.designated_id));
2448 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2449 status.designated_port_id);
2450 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2451 status.bridge_port_id);
2452 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2453 smap_destroy(&smap);
2457 port_refresh_rstp_status(struct port *port)
2459 struct ofproto *ofproto = port->bridge->ofproto;
2460 struct iface *iface;
2461 struct ofproto_port_rstp_status status;
2463 int64_t int_values[3];
2466 if (port_is_synthetic(port)) {
2470 /* RSTP doesn't currently support bonds. */
2471 if (!list_is_singleton(&port->ifaces)) {
2472 ovsrec_port_set_rstp_status(port->cfg, NULL);
2476 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2477 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2481 if (!status.enabled) {
2482 ovsrec_port_set_rstp_status(port->cfg, NULL);
2483 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2486 /* Set Status column. */
2489 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2491 smap_add_format(&smap, "rstp_port_role", "%s",
2492 rstp_port_role_name(status.role));
2493 smap_add_format(&smap, "rstp_port_state", "%s",
2494 rstp_state_name(status.state));
2496 ovsrec_port_set_rstp_status(port->cfg, &smap);
2497 smap_destroy(&smap);
2499 /* Set Statistics column. */
2500 keys[0] = "rstp_tx_count";
2501 int_values[0] = status.tx_count;
2502 keys[1] = "rstp_rx_count";
2503 int_values[1] = status.rx_count;
2504 keys[2] = "rstp_uptime";
2505 int_values[2] = status.uptime;
2506 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2507 ARRAY_SIZE(int_values));
2511 port_refresh_bond_status(struct port *port, bool force_update)
2515 /* Return if port is not a bond */
2516 if (list_is_singleton(&port->ifaces)) {
2520 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2524 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2525 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2531 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2533 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2537 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2539 bool enable = enable_system_stats(cfg);
2541 system_stats_enable(enable);
2543 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2548 run_system_stats(void)
2550 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2553 stats = system_stats_run();
2555 struct ovsdb_idl_txn *txn;
2556 struct ovsdb_datum datum;
2558 txn = ovsdb_idl_txn_create(idl);
2559 ovsdb_datum_from_smap(&datum, stats);
2560 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2562 ovsdb_idl_txn_commit(txn);
2563 ovsdb_idl_txn_destroy(txn);
2570 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2573 case OFPCR12_ROLE_EQUAL:
2575 case OFPCR12_ROLE_MASTER:
2577 case OFPCR12_ROLE_SLAVE:
2579 case OFPCR12_ROLE_NOCHANGE:
2581 return "*** INVALID ROLE ***";
2586 refresh_controller_status(void)
2590 const struct ovsrec_controller *cfg;
2594 /* Accumulate status for controllers on all bridges. */
2595 HMAP_FOR_EACH (br, node, &all_bridges) {
2596 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2599 /* Update each controller in the database with current status. */
2600 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2601 struct ofproto_controller_info *cinfo =
2602 shash_find_data(&info, cfg->target);
2605 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2606 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2608 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2610 ovsrec_controller_set_is_connected(cfg, false);
2611 ovsrec_controller_set_role(cfg, NULL);
2612 ovsrec_controller_set_status(cfg, NULL);
2616 ofproto_free_ofproto_controller_info(&info);
2619 /* Update interface and mirror statistics if necessary. */
2621 run_stats_update(void)
2623 static struct ovsdb_idl_txn *stats_txn;
2624 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2631 /* Statistics update interval should always be greater than or equal to
2633 stats_interval = MAX(smap_get_int(&cfg->other_config,
2634 "stats-update-interval",
2636 if (stats_timer_interval != stats_interval) {
2637 stats_timer_interval = stats_interval;
2638 stats_timer = LLONG_MIN;
2641 if (time_msec() >= stats_timer) {
2642 enum ovsdb_idl_txn_status status;
2644 /* Rate limit the update. Do not start a new update if the
2645 * previous one is not done. */
2649 stats_txn = ovsdb_idl_txn_create(idl);
2650 HMAP_FOR_EACH (br, node, &all_bridges) {
2654 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2655 struct iface *iface;
2657 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2658 iface_refresh_stats(iface);
2660 port_refresh_stp_stats(port);
2662 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2663 mirror_refresh_stats(m);
2666 refresh_controller_status();
2669 status = ovsdb_idl_txn_commit(stats_txn);
2670 if (status != TXN_INCOMPLETE) {
2671 stats_timer = time_msec() + stats_timer_interval;
2672 ovsdb_idl_txn_destroy(stats_txn);
2678 /* Update bridge/port/interface status if necessary. */
2680 run_status_update(void)
2685 /* Rate limit the update. Do not start a new update if the
2686 * previous one is not done. */
2687 seq = seq_read(connectivity_seq_get());
2688 if (seq != connectivity_seqno || status_txn_try_again) {
2691 connectivity_seqno = seq;
2692 status_txn = ovsdb_idl_txn_create(idl);
2693 HMAP_FOR_EACH (br, node, &all_bridges) {
2696 br_refresh_stp_status(br);
2697 br_refresh_rstp_status(br);
2698 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2699 struct iface *iface;
2701 port_refresh_stp_status(port);
2702 port_refresh_rstp_status(port);
2703 port_refresh_bond_status(port, status_txn_try_again);
2704 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2705 iface_refresh_netdev_status(iface);
2706 iface_refresh_ofproto_status(iface);
2713 /* Commit the transaction and get the status. If the transaction finishes,
2714 * then destroy the transaction. Otherwise, keep it so that we can check
2715 * progress the next time that this function is called. */
2717 enum ovsdb_idl_txn_status status;
2719 status = ovsdb_idl_txn_commit(status_txn);
2720 if (status != TXN_INCOMPLETE) {
2721 ovsdb_idl_txn_destroy(status_txn);
2724 /* Sets the 'status_txn_try_again' if the transaction fails. */
2725 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2726 status_txn_try_again = false;
2728 status_txn_try_again = true;
2735 status_update_wait(void)
2737 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2738 * transaction to complete. If the status update to database needs to be
2739 * run again (transaction fails), registers a timeout in
2740 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2741 * sequence number. */
2743 ovsdb_idl_txn_wait(status_txn);
2744 } else if (status_txn_try_again) {
2745 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2747 seq_wait(connectivity_seq_get(), connectivity_seqno);
2758 /* Let each datapath type do the work that it needs to do. */
2760 ofproto_enumerate_types(&types);
2761 SSET_FOR_EACH (type, &types) {
2762 ofproto_type_run(type);
2764 sset_destroy(&types);
2766 /* Let each bridge do the work that it needs to do. */
2767 HMAP_FOR_EACH (br, node, &all_bridges) {
2768 ofproto_run(br->ofproto);
2775 static struct ovsrec_open_vswitch null_cfg;
2776 const struct ovsrec_open_vswitch *cfg;
2778 bool vlan_splinters_changed;
2780 ovsrec_open_vswitch_init(&null_cfg);
2784 if (ovsdb_idl_is_lock_contended(idl)) {
2785 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2786 struct bridge *br, *next_br;
2788 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2789 "disabling this process (pid %ld) until it goes away",
2790 (long int) getpid());
2792 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2795 /* Since we will not be running system_stats_run() in this process
2796 * with the current situation of multiple ovs-vswitchd daemons,
2797 * disable system stats collection. */
2798 system_stats_enable(false);
2799 /* This prevents the process from constantly waking up on
2800 * connectivity seq. */
2801 connectivity_seqno = seq_read(connectivity_seq_get());
2803 } else if (!ovsdb_idl_has_lock(idl)) {
2806 cfg = ovsrec_open_vswitch_first(idl);
2808 /* Initialize the ofproto library. This only needs to run once, but
2809 * it must be done after the configuration is set. If the
2810 * initialization has already occurred, bridge_init_ofproto()
2811 * returns immediately. */
2812 bridge_init_ofproto(cfg);
2814 /* Once the value of flow-restore-wait is false, we no longer should
2815 * check its value from the database. */
2816 if (cfg && ofproto_get_flow_restore_wait()) {
2817 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2818 "flow-restore-wait", false));
2823 /* Re-configure SSL. We do this on every trip through the main loop,
2824 * instead of just when the database changes, because the contents of the
2825 * key and certificate files can change without the database changing.
2827 * We do this before bridge_reconfigure() because that function might
2828 * initiate SSL connections and thus requires SSL to be configured. */
2829 if (cfg && cfg->ssl) {
2830 const struct ovsrec_ssl *ssl = cfg->ssl;
2832 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2833 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2836 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2837 * usage has changed. */
2838 vlan_splinters_changed = false;
2839 if (vlan_splinters_enabled_anywhere) {
2842 HMAP_FOR_EACH (br, node, &all_bridges) {
2843 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2844 vlan_splinters_changed = true;
2850 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2851 struct ovsdb_idl_txn *txn;
2853 idl_seqno = ovsdb_idl_get_seqno(idl);
2854 txn = ovsdb_idl_txn_create(idl);
2855 bridge_reconfigure(cfg ? cfg : &null_cfg);
2858 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2861 /* If we are completing our initial configuration for this run
2862 * of ovs-vswitchd, then keep the transaction around to monitor
2863 * it for completion. */
2864 if (initial_config_done) {
2865 /* Always sets the 'status_txn_try_again' to check again,
2866 * in case that this transaction fails. */
2867 status_txn_try_again = true;
2868 ovsdb_idl_txn_commit(txn);
2869 ovsdb_idl_txn_destroy(txn);
2871 initial_config_done = true;
2872 daemonize_txn = txn;
2876 if (daemonize_txn) {
2877 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2878 if (status != TXN_INCOMPLETE) {
2879 ovsdb_idl_txn_destroy(daemonize_txn);
2880 daemonize_txn = NULL;
2882 /* ovs-vswitchd has completed initialization, so allow the
2883 * process that forked us to exit successfully. */
2884 daemonize_complete();
2886 vlog_enable_async();
2888 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2893 run_status_update();
2903 ovsdb_idl_wait(idl);
2904 if (daemonize_txn) {
2905 ovsdb_idl_txn_wait(daemonize_txn);
2909 ofproto_enumerate_types(&types);
2910 SSET_FOR_EACH (type, &types) {
2911 ofproto_type_wait(type);
2913 sset_destroy(&types);
2915 if (!hmap_is_empty(&all_bridges)) {
2918 HMAP_FOR_EACH (br, node, &all_bridges) {
2919 ofproto_wait(br->ofproto);
2922 poll_timer_wait_until(stats_timer);
2925 status_update_wait();
2926 system_stats_wait();
2929 /* Adds some memory usage statistics for bridges into 'usage', for use with
2930 * memory_report(). */
2932 bridge_get_memory_usage(struct simap *usage)
2939 ofproto_enumerate_types(&types);
2940 SSET_FOR_EACH (type, &types) {
2941 ofproto_type_get_memory_usage(type, usage);
2943 sset_destroy(&types);
2945 HMAP_FOR_EACH (br, node, &all_bridges) {
2946 ofproto_get_memory_usage(br->ofproto, usage);
2950 /* QoS unixctl user interface functions. */
2952 struct qos_unixctl_show_cbdata {
2954 struct iface *iface;
2958 qos_unixctl_show_queue(unsigned int queue_id,
2959 const struct smap *details,
2960 struct iface *iface,
2963 struct netdev_queue_stats stats;
2964 struct smap_node *node;
2967 ds_put_cstr(ds, "\n");
2969 ds_put_format(ds, "Queue %u:\n", queue_id);
2971 ds_put_cstr(ds, "Default:\n");
2974 SMAP_FOR_EACH (node, details) {
2975 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2978 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2980 if (stats.tx_packets != UINT64_MAX) {
2981 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2984 if (stats.tx_bytes != UINT64_MAX) {
2985 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2988 if (stats.tx_errors != UINT64_MAX) {
2989 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2992 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2993 queue_id, ovs_strerror(error));
2998 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2999 const char *argv[], void *aux OVS_UNUSED)
3001 struct ds ds = DS_EMPTY_INITIALIZER;
3002 struct smap smap = SMAP_INITIALIZER(&smap);
3003 struct iface *iface;
3005 struct smap_node *node;
3007 iface = iface_find(argv[1]);
3009 unixctl_command_reply_error(conn, "no such interface");
3013 netdev_get_qos(iface->netdev, &type, &smap);
3015 if (*type != '\0') {
3016 struct netdev_queue_dump dump;
3017 struct smap details;
3018 unsigned int queue_id;
3020 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3022 SMAP_FOR_EACH (node, &smap) {
3023 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3026 smap_init(&details);
3027 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3028 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3030 smap_destroy(&details);
3032 unixctl_command_reply(conn, ds_cstr(&ds));
3034 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3035 unixctl_command_reply_error(conn, ds_cstr(&ds));
3038 smap_destroy(&smap);
3042 /* Bridge reconfiguration functions. */
3044 bridge_create(const struct ovsrec_bridge *br_cfg)
3048 ovs_assert(!bridge_lookup(br_cfg->name));
3049 br = xzalloc(sizeof *br);
3051 br->name = xstrdup(br_cfg->name);
3052 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3055 /* Derive the default Ethernet address from the bridge's UUID. This should
3056 * be unique and it will be stable between ovs-vswitchd runs. */
3057 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3058 eth_addr_mark_random(br->default_ea);
3060 hmap_init(&br->ports);
3061 hmap_init(&br->ifaces);
3062 hmap_init(&br->iface_by_name);
3063 hmap_init(&br->mirrors);
3065 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3069 bridge_destroy(struct bridge *br)
3072 struct mirror *mirror, *next_mirror;
3073 struct port *port, *next_port;
3075 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3078 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3079 mirror_destroy(mirror);
3082 hmap_remove(&all_bridges, &br->node);
3083 ofproto_destroy(br->ofproto);
3084 hmap_destroy(&br->ifaces);
3085 hmap_destroy(&br->ports);
3086 hmap_destroy(&br->iface_by_name);
3087 hmap_destroy(&br->mirrors);
3094 static struct bridge *
3095 bridge_lookup(const char *name)
3099 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3100 if (!strcmp(br->name, name)) {
3107 /* Handle requests for a listing of all flows known by the OpenFlow
3108 * stack, including those normally hidden. */
3110 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3111 const char *argv[], void *aux OVS_UNUSED)
3116 br = bridge_lookup(argv[1]);
3118 unixctl_command_reply_error(conn, "Unknown bridge");
3123 ofproto_get_all_flows(br->ofproto, &results);
3125 unixctl_command_reply(conn, ds_cstr(&results));
3126 ds_destroy(&results);
3129 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3130 * connections and reconnect. If BRIDGE is not specified, then all bridges
3131 * drop their controller connections and reconnect. */
3133 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3134 const char *argv[], void *aux OVS_UNUSED)
3138 br = bridge_lookup(argv[1]);
3140 unixctl_command_reply_error(conn, "Unknown bridge");
3143 ofproto_reconnect_controllers(br->ofproto);
3145 HMAP_FOR_EACH (br, node, &all_bridges) {
3146 ofproto_reconnect_controllers(br->ofproto);
3149 unixctl_command_reply(conn, NULL);
3153 bridge_get_controllers(const struct bridge *br,
3154 struct ovsrec_controller ***controllersp)
3156 struct ovsrec_controller **controllers;
3157 size_t n_controllers;
3159 controllers = br->cfg->controller;
3160 n_controllers = br->cfg->n_controller;
3162 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3168 *controllersp = controllers;
3170 return n_controllers;
3174 bridge_collect_wanted_ports(struct bridge *br,
3175 const unsigned long int *splinter_vlans,
3176 struct shash *wanted_ports)
3180 shash_init(wanted_ports);
3182 for (i = 0; i < br->cfg->n_ports; i++) {
3183 const char *name = br->cfg->ports[i]->name;
3184 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3185 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3189 if (bridge_get_controllers(br, NULL)
3190 && !shash_find(wanted_ports, br->name)) {
3191 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3192 br->name, br->name);
3194 ovsrec_interface_init(&br->synth_local_iface);
3195 ovsrec_port_init(&br->synth_local_port);
3197 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3198 br->synth_local_port.n_interfaces = 1;
3199 br->synth_local_port.name = br->name;
3201 br->synth_local_iface.name = br->name;
3202 br->synth_local_iface.type = "internal";
3204 br->synth_local_ifacep = &br->synth_local_iface;
3206 shash_add(wanted_ports, br->name, &br->synth_local_port);
3209 if (splinter_vlans) {
3210 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3214 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3215 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3216 * 'br' needs to complete its configuration. */
3218 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3220 struct shash_node *port_node;
3221 struct port *port, *next;
3223 /* Get rid of deleted ports.
3224 * Get rid of deleted interfaces on ports that still exist. */
3225 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3226 port->cfg = shash_find_data(wanted_ports, port->name);
3230 port_del_ifaces(port);
3234 /* Update iface->cfg and iface->type in interfaces that still exist. */
3235 SHASH_FOR_EACH (port_node, wanted_ports) {
3236 const struct ovsrec_port *port = port_node->data;
3239 for (i = 0; i < port->n_interfaces; i++) {
3240 const struct ovsrec_interface *cfg = port->interfaces[i];
3241 struct iface *iface = iface_lookup(br, cfg->name);
3242 const char *type = iface_get_type(cfg, br->cfg);
3247 } else if (!strcmp(type, "null")) {
3248 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3249 " may be removed in February 2013. Please email"
3250 " dev@openvswitch.org with concerns.",
3253 /* We will add new interfaces later. */
3259 /* Initializes 'oc' appropriately as a management service controller for
3262 * The caller must free oc->target when it is no longer needed. */
3264 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3265 struct ofproto_controller *oc)
3267 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3268 oc->max_backoff = 0;
3269 oc->probe_interval = 60;
3270 oc->band = OFPROTO_OUT_OF_BAND;
3272 oc->burst_limit = 0;
3273 oc->enable_async_msgs = true;
3277 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3279 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3280 struct ofproto_controller *oc)
3284 oc->target = c->target;
3285 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3286 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3287 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3288 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3289 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3290 oc->burst_limit = (c->controller_burst_limit
3291 ? *c->controller_burst_limit : 0);
3292 oc->enable_async_msgs = (!c->enable_async_messages
3293 || *c->enable_async_messages);
3294 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3295 if (dscp < 0 || dscp > 63) {
3296 dscp = DSCP_DEFAULT;
3301 /* Configures the IP stack for 'br''s local interface properly according to the
3302 * configuration in 'c'. */
3304 bridge_configure_local_iface_netdev(struct bridge *br,
3305 struct ovsrec_controller *c)
3307 struct netdev *netdev;
3308 struct in_addr mask, gateway;
3310 struct iface *local_iface;
3313 /* If there's no local interface or no IP address, give up. */
3314 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3315 if (!local_iface || !c->local_ip
3316 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3320 /* Bring up the local interface. */
3321 netdev = local_iface->netdev;
3322 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3324 /* Configure the IP address and netmask. */
3325 if (!c->local_netmask
3326 || !inet_pton(AF_INET, c->local_netmask, &mask)
3328 mask.s_addr = guess_netmask(ip.s_addr);
3330 if (!netdev_set_in4(netdev, ip, mask)) {
3331 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3332 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3335 /* Configure the default gateway. */
3336 if (c->local_gateway
3337 && inet_pton(AF_INET, c->local_gateway, &gateway)
3338 && gateway.s_addr) {
3339 if (!netdev_add_router(netdev, gateway)) {
3340 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3341 br->name, IP_ARGS(gateway.s_addr));
3346 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3347 * in either string are treated as equal to any number of slashes in the other,
3348 * e.g. "x///y" is equal to "x/y".
3350 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3351 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3352 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3353 * 'b' against a prefix of 'a'.
3356 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3358 const char *b_start = b;
3360 if (b - b_start >= b_stoplen) {
3362 } else if (*a != *b) {
3364 } else if (*a == '/') {
3365 a += strspn(a, "/");
3366 b += strspn(b, "/");
3367 } else if (*a == '\0') {
3377 bridge_configure_remotes(struct bridge *br,
3378 const struct sockaddr_in *managers, size_t n_managers)
3380 bool disable_in_band;
3382 struct ovsrec_controller **controllers;
3383 size_t n_controllers;
3385 enum ofproto_fail_mode fail_mode;
3387 struct ofproto_controller *ocs;
3391 /* Check if we should disable in-band control on this bridge. */
3392 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3395 /* Set OpenFlow queue ID for in-band control. */
3396 ofproto_set_in_band_queue(br->ofproto,
3397 smap_get_int(&br->cfg->other_config,
3398 "in-band-queue", -1));
3400 if (disable_in_band) {
3401 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3403 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3406 n_controllers = bridge_get_controllers(br, &controllers);
3408 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3411 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3412 for (i = 0; i < n_controllers; i++) {
3413 struct ovsrec_controller *c = controllers[i];
3415 if (!strncmp(c->target, "punix:", 6)
3416 || !strncmp(c->target, "unix:", 5)) {
3417 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3420 if (!strncmp(c->target, "unix:", 5)) {
3421 /* Connect to a listening socket */
3422 whitelist = xasprintf("unix:%s/", ovs_rundir());
3423 if (strchr(c->target, '/') &&
3424 !equal_pathnames(c->target, whitelist,
3425 strlen(whitelist))) {
3426 /* Absolute path specified, but not in ovs_rundir */
3427 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3428 "controller \"%s\" due to possibility for "
3429 "remote exploit. Instead, specify socket "
3430 "in whitelisted \"%s\" or connect to "
3431 "\"unix:%s/%s.mgmt\" (which is always "
3432 "available without special configuration).",
3433 br->name, c->target, whitelist,
3434 ovs_rundir(), br->name);
3439 whitelist = xasprintf("punix:%s/%s.controller",
3440 ovs_rundir(), br->name);
3441 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3442 /* Prevent remote ovsdb-server users from accessing
3443 * arbitrary Unix domain sockets and overwriting arbitrary
3445 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3446 "controller \"%s\" due to possibility of "
3447 "overwriting local files. Instead, specify "
3448 "whitelisted \"%s\" or connect to "
3449 "\"unix:%s/%s.mgmt\" (which is always "
3450 "available without special configuration).",
3451 br->name, c->target, whitelist,
3452 ovs_rundir(), br->name);
3461 bridge_configure_local_iface_netdev(br, c);
3462 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3463 if (disable_in_band) {
3464 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3469 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3470 bridge_get_allowed_versions(br));
3471 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3474 /* Set the fail-mode. */
3475 fail_mode = !br->cfg->fail_mode
3476 || !strcmp(br->cfg->fail_mode, "standalone")
3477 ? OFPROTO_FAIL_STANDALONE
3478 : OFPROTO_FAIL_SECURE;
3479 ofproto_set_fail_mode(br->ofproto, fail_mode);
3481 /* Configure OpenFlow controller connection snooping. */
3482 if (!ofproto_has_snoops(br->ofproto)) {
3486 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3487 ovs_rundir(), br->name));
3488 ofproto_set_snoops(br->ofproto, &snoops);
3489 sset_destroy(&snoops);
3494 bridge_configure_tables(struct bridge *br)
3496 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3500 n_tables = ofproto_get_n_tables(br->ofproto);
3502 for (i = 0; i < n_tables; i++) {
3503 struct ofproto_table_settings s;
3504 bool use_default_prefixes = true;
3507 s.max_flows = UINT_MAX;
3510 s.n_prefix_fields = 0;
3511 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3513 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3514 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3517 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3518 s.max_flows = *cfg->flow_limit;
3520 if (cfg->overflow_policy
3521 && !strcmp(cfg->overflow_policy, "evict")) {
3523 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3524 for (k = 0; k < cfg->n_groups; k++) {
3525 const char *string = cfg->groups[k];
3528 msg = mf_parse_subfield__(&s.groups[k], &string);
3530 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3531 "'groups' (%s)", br->name, i, msg);
3533 } else if (*string) {
3534 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3535 "element '%s' contains trailing garbage",
3536 br->name, i, cfg->groups[k]);
3542 /* Prefix lookup fields. */
3543 s.n_prefix_fields = 0;
3544 for (k = 0; k < cfg->n_prefixes; k++) {
3545 const char *name = cfg->prefixes[k];
3546 const struct mf_field *mf;
3548 if (strcmp(name, "none") == 0) {
3549 use_default_prefixes = false;
3550 s.n_prefix_fields = 0;
3553 mf = mf_from_name(name);
3555 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3559 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3560 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3561 "%s", br->name, name);
3564 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3565 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3566 "field not used: %s", br->name, name);
3569 use_default_prefixes = false;
3570 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3573 if (use_default_prefixes) {
3574 /* Use default values. */
3575 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3576 memcpy(s.prefix_fields, default_prefix_fields,
3577 sizeof default_prefix_fields);
3580 struct ds ds = DS_EMPTY_INITIALIZER;
3581 for (k = 0; k < s.n_prefix_fields; k++) {
3583 ds_put_char(&ds, ',');
3585 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3587 if (s.n_prefix_fields == 0) {
3588 ds_put_cstr(&ds, "none");
3590 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3591 br->name, i, ds_cstr(&ds));
3595 ofproto_configure_table(br->ofproto, i, &s);
3599 for (; j < br->cfg->n_flow_tables; j++) {
3600 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3601 "%"PRId64" not supported by this datapath", br->name,
3602 br->cfg->key_flow_tables[j]);
3607 bridge_configure_dp_desc(struct bridge *br)
3609 ofproto_set_dp_desc(br->ofproto,
3610 smap_get(&br->cfg->other_config, "dp-desc"));
3613 /* Port functions. */
3615 static struct port *
3616 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3620 port = xzalloc(sizeof *port);
3622 port->name = xstrdup(cfg->name);
3624 list_init(&port->ifaces);
3626 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3630 /* Deletes interfaces from 'port' that are no longer configured for it. */
3632 port_del_ifaces(struct port *port)
3634 struct iface *iface, *next;
3635 struct sset new_ifaces;
3638 /* Collect list of new interfaces. */
3639 sset_init(&new_ifaces);
3640 for (i = 0; i < port->cfg->n_interfaces; i++) {
3641 const char *name = port->cfg->interfaces[i]->name;
3642 const char *type = port->cfg->interfaces[i]->type;
3643 if (strcmp(type, "null")) {
3644 sset_add(&new_ifaces, name);
3648 /* Get rid of deleted interfaces. */
3649 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3650 if (!sset_contains(&new_ifaces, iface->name)) {
3651 iface_destroy(iface);
3655 sset_destroy(&new_ifaces);
3659 port_destroy(struct port *port)
3662 struct bridge *br = port->bridge;
3663 struct iface *iface, *next;
3666 ofproto_bundle_unregister(br->ofproto, port);
3669 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3670 iface_destroy__(iface);
3673 hmap_remove(&br->ports, &port->hmap_node);
3679 static struct port *
3680 port_lookup(const struct bridge *br, const char *name)
3684 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3686 if (!strcmp(port->name, name)) {
3694 enable_lacp(struct port *port, bool *activep)
3696 if (!port->cfg->lacp) {
3697 /* XXX when LACP implementation has been sufficiently tested, enable by
3698 * default and make active on bonded ports. */
3700 } else if (!strcmp(port->cfg->lacp, "off")) {
3702 } else if (!strcmp(port->cfg->lacp, "active")) {
3705 } else if (!strcmp(port->cfg->lacp, "passive")) {
3709 VLOG_WARN("port %s: unknown LACP mode %s",
3710 port->name, port->cfg->lacp);
3715 static struct lacp_settings *
3716 port_configure_lacp(struct port *port, struct lacp_settings *s)
3718 const char *lacp_time, *system_id;
3721 if (!enable_lacp(port, &s->active)) {
3725 s->name = port->name;
3727 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3729 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3730 ETH_ADDR_SCAN_ARGS(s->id))) {
3731 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3732 " address.", port->name, system_id);
3736 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3739 if (eth_addr_is_zero(s->id)) {
3740 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3744 /* Prefer bondable links if unspecified. */
3745 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3747 s->priority = (priority > 0 && priority <= UINT16_MAX
3749 : UINT16_MAX - !list_is_short(&port->ifaces));
3751 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3752 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3754 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3755 "lacp-fallback-ab", false);
3761 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3763 int priority, portid, key;
3765 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3766 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3768 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3770 if (portid <= 0 || portid > UINT16_MAX) {
3771 portid = ofp_to_u16(iface->ofp_port);
3774 if (priority <= 0 || priority > UINT16_MAX) {
3775 priority = UINT16_MAX;
3778 if (key < 0 || key > UINT16_MAX) {
3782 s->name = iface->name;
3784 s->priority = priority;
3789 port_configure_bond(struct port *port, struct bond_settings *s)
3791 const char *detect_s;
3792 struct iface *iface;
3794 int miimon_interval;
3796 s->name = port->name;
3798 if (port->cfg->bond_mode) {
3799 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3800 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3801 port->name, port->cfg->bond_mode,
3802 bond_mode_to_string(s->balance));
3805 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3807 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3808 * active-backup. At some point we should remove this warning. */
3809 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3810 " in previous versions, the default bond_mode was"
3811 " balance-slb", port->name,
3812 bond_mode_to_string(s->balance));
3814 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3815 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3816 "please use another bond type or disable flood_vlans",
3820 miimon_interval = smap_get_int(&port->cfg->other_config,
3821 "bond-miimon-interval", 0);
3822 if (miimon_interval <= 0) {
3823 miimon_interval = 200;
3826 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3827 if (!detect_s || !strcmp(detect_s, "carrier")) {
3828 miimon_interval = 0;
3829 } else if (strcmp(detect_s, "miimon")) {
3830 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3831 "defaulting to carrier", port->name, detect_s);
3832 miimon_interval = 0;
3835 s->up_delay = MAX(0, port->cfg->bond_updelay);
3836 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3837 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3838 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3839 "bond-rebalance-interval", 10000);
3840 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3841 s->rebalance_interval = 1000;
3844 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3845 "lacp-fallback-ab", false);
3847 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3848 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3851 mac_s = port->cfg->bond_active_slave;
3852 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
3853 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
3854 /* OVSDB did not store the last active interface */
3855 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
3859 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3860 * instead of obtaining it from the database. */
3862 port_is_synthetic(const struct port *port)
3864 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3867 /* Interface functions. */
3870 iface_is_internal(const struct ovsrec_interface *iface,
3871 const struct ovsrec_bridge *br)
3873 /* The local port and "internal" ports are always "internal". */
3874 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3877 /* Returns the correct network device type for interface 'iface' in bridge
3880 iface_get_type(const struct ovsrec_interface *iface,
3881 const struct ovsrec_bridge *br)
3885 /* The local port always has type "internal". Other ports take
3886 * their type from the database and default to "system" if none is
3888 if (iface_is_internal(iface, br)) {
3891 type = iface->type[0] ? iface->type : "system";
3894 return ofproto_port_open_type(br->datapath_type, type);
3898 iface_destroy__(struct iface *iface)
3901 struct port *port = iface->port;
3902 struct bridge *br = port->bridge;
3904 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3905 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3908 if (iface->ofp_port != OFPP_NONE) {
3909 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3912 list_remove(&iface->port_elem);
3913 hmap_remove(&br->iface_by_name, &iface->name_node);
3915 /* The user is changing configuration here, so netdev_remove needs to be
3916 * used as opposed to netdev_close */
3917 netdev_remove(iface->netdev);
3925 iface_destroy(struct iface *iface)
3928 struct port *port = iface->port;
3930 iface_destroy__(iface);
3931 if (list_is_empty(&port->ifaces)) {
3937 static struct iface *
3938 iface_lookup(const struct bridge *br, const char *name)
3940 struct iface *iface;
3942 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3943 &br->iface_by_name) {
3944 if (!strcmp(iface->name, name)) {
3952 static struct iface *
3953 iface_find(const char *name)
3955 const struct bridge *br;
3957 HMAP_FOR_EACH (br, node, &all_bridges) {
3958 struct iface *iface = iface_lookup(br, name);
3967 static struct iface *
3968 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3970 struct iface *iface;
3972 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3974 if (iface->ofp_port == ofp_port) {
3981 /* Set Ethernet address of 'iface', if one is specified in the configuration
3984 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3986 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3987 struct iface *hw_addr_iface;
3989 if (strcmp(iface->type, "internal")) {
3993 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3995 } else if (port->cfg->fake_bridge) {
3996 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3997 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4002 if (iface->ofp_port == OFPP_LOCAL) {
4003 VLOG_ERR("interface %s: ignoring mac in Interface record "
4004 "(use Bridge record to set local port's mac)",
4006 } else if (eth_addr_is_multicast(mac)) {
4007 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4010 int error = netdev_set_etheraddr(iface->netdev, mac);
4012 VLOG_ERR("interface %s: setting MAC failed (%s)",
4013 iface->name, ovs_strerror(error));
4019 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4021 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4023 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4024 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4025 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4029 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4030 * sets the "ofport" field to -1.
4032 * This is appropriate when 'if_cfg''s interface cannot be created or is
4033 * otherwise invalid. */
4035 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4037 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4038 iface_set_ofport(if_cfg, OFPP_NONE);
4039 ovsrec_interface_set_error(if_cfg, errp);
4040 ovsrec_interface_set_status(if_cfg, NULL);
4041 ovsrec_interface_set_admin_state(if_cfg, NULL);
4042 ovsrec_interface_set_duplex(if_cfg, NULL);
4043 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4044 ovsrec_interface_set_link_state(if_cfg, NULL);
4045 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4046 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4047 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4048 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4049 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4050 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4051 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4052 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4057 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4059 union ovsdb_atom atom;
4061 atom.integer = target;
4062 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4066 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4068 struct ofpbuf queues_buf;
4070 ofpbuf_init(&queues_buf, 0);
4072 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
4073 netdev_set_qos(iface->netdev, NULL, NULL);
4075 const struct ovsdb_datum *queues;
4076 struct netdev_queue_dump dump;
4077 unsigned int queue_id;
4078 struct smap details;
4082 /* Configure top-level Qos for 'iface'. */
4083 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4085 /* Deconfigure queues that were deleted. */
4086 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4088 smap_init(&details);
4089 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4090 if (!queue_ids_include(queues, queue_id)) {
4091 netdev_delete_queue(iface->netdev, queue_id);
4094 smap_destroy(&details);
4096 /* Configure queues for 'iface'. */
4098 for (i = 0; i < qos->n_queues; i++) {
4099 const struct ovsrec_queue *queue = qos->value_queues[i];
4100 unsigned int queue_id = qos->key_queues[i];
4102 if (queue_id == 0) {
4106 if (queue->n_dscp == 1) {
4107 struct ofproto_port_queue *port_queue;
4109 port_queue = ofpbuf_put_uninit(&queues_buf,
4110 sizeof *port_queue);
4111 port_queue->queue = queue_id;
4112 port_queue->dscp = queue->dscp[0];
4115 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4118 struct smap details;
4120 smap_init(&details);
4121 netdev_set_queue(iface->netdev, 0, &details);
4122 smap_destroy(&details);
4126 if (iface->ofp_port != OFPP_NONE) {
4127 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
4128 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
4130 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4131 port_queues, n_queues);
4134 netdev_set_policing(iface->netdev,
4135 iface->cfg->ingress_policing_rate,
4136 iface->cfg->ingress_policing_burst);
4138 ofpbuf_uninit(&queues_buf);
4142 iface_configure_cfm(struct iface *iface)
4144 const struct ovsrec_interface *cfg = iface->cfg;
4145 const char *opstate_str;
4146 const char *cfm_ccm_vlan;
4147 struct cfm_settings s;
4148 struct smap netdev_args;
4150 if (!cfg->n_cfm_mpid) {
4151 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4155 s.check_tnl_key = false;
4156 smap_init(&netdev_args);
4157 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4158 const char *key = smap_get(&netdev_args, "key");
4159 const char *in_key = smap_get(&netdev_args, "in_key");
4161 s.check_tnl_key = (key && !strcmp(key, "flow"))
4162 || (in_key && !strcmp(in_key, "flow"));
4164 smap_destroy(&netdev_args);
4166 s.mpid = *cfg->cfm_mpid;
4167 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4168 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4169 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4171 if (s.interval <= 0) {
4175 if (!cfm_ccm_vlan) {
4177 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4178 s.ccm_vlan = CFM_RANDOM_VLAN;
4180 s.ccm_vlan = atoi(cfm_ccm_vlan);
4181 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4186 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4188 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4190 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4191 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4193 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4196 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4197 * instead of obtaining it from the database. */
4199 iface_is_synthetic(const struct iface *iface)
4201 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4205 iface_validate_ofport__(size_t n, int64_t *ofport)
4207 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4208 ? u16_to_ofp(*ofport)
4213 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4215 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4219 iface_pick_ofport(const struct ovsrec_interface *cfg)
4221 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4222 return (requested_ofport != OFPP_NONE
4224 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4227 /* Port mirroring. */
4229 static struct mirror *
4230 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4234 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4235 if (uuid_equals(uuid, &m->uuid)) {
4243 bridge_configure_mirrors(struct bridge *br)
4245 const struct ovsdb_datum *mc;
4246 unsigned long *flood_vlans;
4247 struct mirror *m, *next;
4250 /* Get rid of deleted mirrors. */
4251 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4252 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4253 union ovsdb_atom atom;
4255 atom.uuid = m->uuid;
4256 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4261 /* Add new mirrors and reconfigure existing ones. */
4262 for (i = 0; i < br->cfg->n_mirrors; i++) {
4263 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4264 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4266 m = mirror_create(br, cfg);
4269 if (!mirror_configure(m)) {
4274 /* Update flooded vlans (for RSPAN). */
4275 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4276 br->cfg->n_flood_vlans);
4277 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4278 bitmap_free(flood_vlans);
4281 static struct mirror *
4282 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4286 m = xzalloc(sizeof *m);
4287 m->uuid = cfg->header_.uuid;
4288 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4290 m->name = xstrdup(cfg->name);
4296 mirror_destroy(struct mirror *m)
4299 struct bridge *br = m->bridge;
4302 ofproto_mirror_unregister(br->ofproto, m);
4305 hmap_remove(&br->mirrors, &m->hmap_node);
4312 mirror_collect_ports(struct mirror *m,
4313 struct ovsrec_port **in_ports, int n_in_ports,
4314 void ***out_portsp, size_t *n_out_portsp)
4316 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4317 size_t n_out_ports = 0;
4320 for (i = 0; i < n_in_ports; i++) {
4321 const char *name = in_ports[i]->name;
4322 struct port *port = port_lookup(m->bridge, name);
4324 out_ports[n_out_ports++] = port;
4326 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4327 "port %s", m->bridge->name, m->name, name);
4330 *out_portsp = out_ports;
4331 *n_out_portsp = n_out_ports;
4335 mirror_configure(struct mirror *m)
4337 const struct ovsrec_mirror *cfg = m->cfg;
4338 struct ofproto_mirror_settings s;
4341 if (strcmp(cfg->name, m->name)) {
4343 m->name = xstrdup(cfg->name);
4347 /* Get output port or VLAN. */
4348 if (cfg->output_port) {
4349 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4350 if (!s.out_bundle) {
4351 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4352 m->bridge->name, m->name);
4355 s.out_vlan = UINT16_MAX;
4357 if (cfg->output_vlan) {
4358 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4359 "output vlan; ignoring output vlan",
4360 m->bridge->name, m->name);
4362 } else if (cfg->output_vlan) {
4363 /* The database should prevent invalid VLAN values. */
4364 s.out_bundle = NULL;
4365 s.out_vlan = *cfg->output_vlan;
4367 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4368 m->bridge->name, m->name);
4372 /* Get port selection. */
4373 if (cfg->select_all) {
4374 size_t n_ports = hmap_count(&m->bridge->ports);
4375 void **ports = xmalloc(n_ports * sizeof *ports);
4380 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4390 /* Get ports, dropping ports that don't exist.
4391 * The IDL ensures that there are no duplicates. */
4392 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4393 &s.srcs, &s.n_srcs);
4394 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4395 &s.dsts, &s.n_dsts);
4398 /* Get VLAN selection. */
4399 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4402 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4405 if (s.srcs != s.dsts) {
4414 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4416 * This is deprecated. It is only for compatibility with broken device drivers
4417 * in old versions of Linux that do not properly support VLANs when VLAN
4418 * devices are not used. When broken device drivers are no longer in
4419 * widespread use, we will delete these interfaces. */
4421 static struct ovsrec_port **recs;
4422 static size_t n_recs, allocated_recs;
4424 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4425 * splinters are reconfigured. */
4427 register_rec(struct ovsrec_port *rec)
4429 if (n_recs >= allocated_recs) {
4430 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4432 recs[n_recs++] = rec;
4435 /* Frees all of the ports registered with register_reg(). */
4437 free_registered_recs(void)
4441 for (i = 0; i < n_recs; i++) {
4442 struct ovsrec_port *port = recs[i];
4445 for (j = 0; j < port->n_interfaces; j++) {
4446 struct ovsrec_interface *iface = port->interfaces[j];
4451 smap_destroy(&port->other_config);
4452 free(port->interfaces);
4460 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4463 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4465 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4469 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4472 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4473 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4474 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4477 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4478 * use, returns NULL.
4480 * Updates 'vlan_splinters_enabled_anywhere'. */
4481 static unsigned long int *
4482 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4484 unsigned long int *splinter_vlans;
4485 struct sset splinter_ifaces;
4486 const char *real_dev_name;
4487 struct shash *real_devs;
4488 struct shash_node *node;
4492 /* Free space allocated for synthesized ports and interfaces, since we're
4493 * in the process of reconstructing all of them. */
4494 free_registered_recs();
4496 splinter_vlans = bitmap_allocate(4096);
4497 sset_init(&splinter_ifaces);
4498 vlan_splinters_enabled_anywhere = false;
4499 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4500 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4503 for (j = 0; j < br_cfg->n_ports; j++) {
4504 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4507 for (k = 0; k < port_cfg->n_interfaces; k++) {
4508 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4510 if (vlan_splinters_is_enabled(iface_cfg)) {
4511 vlan_splinters_enabled_anywhere = true;
4512 sset_add(&splinter_ifaces, iface_cfg->name);
4513 vlan_bitmap_from_array__(port_cfg->trunks,
4519 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4520 bitmap_set1(splinter_vlans, *port_cfg->tag);
4525 if (!vlan_splinters_enabled_anywhere) {
4526 free(splinter_vlans);
4527 sset_destroy(&splinter_ifaces);
4531 HMAP_FOR_EACH (br, node, &all_bridges) {
4533 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4537 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4538 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4539 * device to be created for it. */
4540 bitmap_set0(splinter_vlans, 0);
4541 bitmap_set0(splinter_vlans, 4095);
4543 /* Delete all VLAN devices that we don't need. */
4545 real_devs = vlandev_get_real_devs();
4546 SHASH_FOR_EACH (node, real_devs) {
4547 const struct vlan_real_dev *real_dev = node->data;
4548 const struct vlan_dev *vlan_dev;
4549 bool real_dev_has_splinters;
4551 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4553 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4554 if (!real_dev_has_splinters
4555 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4556 struct netdev *netdev;
4558 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4559 if (!netdev_get_in4(netdev, NULL, NULL) ||
4560 !netdev_get_in6(netdev, NULL)) {
4561 /* It has an IP address configured, so we don't own
4562 * it. Don't delete it. */
4564 vlandev_del(vlan_dev->name);
4566 netdev_close(netdev);
4573 /* Add all VLAN devices that we need. */
4574 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4577 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4578 if (!vlandev_get_name(real_dev_name, vid)) {
4579 vlandev_add(real_dev_name, vid);
4586 sset_destroy(&splinter_ifaces);
4588 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4589 free(splinter_vlans);
4592 return splinter_vlans;
4595 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4598 configure_splinter_port(struct port *port)
4600 struct ofproto *ofproto = port->bridge->ofproto;
4601 ofp_port_t realdev_ofp_port;
4602 const char *realdev_name;
4603 struct iface *vlandev, *realdev;
4605 ofproto_bundle_unregister(port->bridge->ofproto, port);
4607 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4610 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4611 realdev = iface_lookup(port->bridge, realdev_name);
4612 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4614 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4618 static struct ovsrec_port *
4619 synthesize_splinter_port(const char *real_dev_name,
4620 const char *vlan_dev_name, int vid)
4622 struct ovsrec_interface *iface;
4623 struct ovsrec_port *port;
4625 iface = xmalloc(sizeof *iface);
4626 ovsrec_interface_init(iface);
4627 iface->name = xstrdup(vlan_dev_name);
4628 iface->type = "system";
4630 port = xmalloc(sizeof *port);
4631 ovsrec_port_init(port);
4632 port->interfaces = xmemdup(&iface, sizeof iface);
4633 port->n_interfaces = 1;
4634 port->name = xstrdup(vlan_dev_name);
4635 port->vlan_mode = "splinter";
4636 port->tag = xmalloc(sizeof *port->tag);
4639 smap_add(&port->other_config, "realdev", real_dev_name);
4645 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4646 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4647 * 1-bit in the 'splinter_vlans' bitmap. */
4649 add_vlan_splinter_ports(struct bridge *br,
4650 const unsigned long int *splinter_vlans,
4651 struct shash *ports)
4655 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4656 * we're modifying 'ports'. */
4657 for (i = 0; i < br->cfg->n_ports; i++) {
4658 const char *name = br->cfg->ports[i]->name;
4659 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4662 for (j = 0; j < port_cfg->n_interfaces; j++) {
4663 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4665 if (vlan_splinters_is_enabled(iface_cfg)) {
4666 const char *real_dev_name;
4669 real_dev_name = iface_cfg->name;
4670 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4671 const char *vlan_dev_name;
4673 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4675 && !shash_find(ports, vlan_dev_name)) {
4676 shash_add(ports, vlan_dev_name,
4677 synthesize_splinter_port(
4678 real_dev_name, vlan_dev_name, vid));
4687 mirror_refresh_stats(struct mirror *m)
4689 struct ofproto *ofproto = m->bridge->ofproto;
4690 uint64_t tx_packets, tx_bytes;
4693 size_t stat_cnt = 0;
4695 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4696 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4700 if (tx_packets != UINT64_MAX) {
4701 keys[stat_cnt] = "tx_packets";
4702 values[stat_cnt] = tx_packets;
4705 if (tx_bytes != UINT64_MAX) {
4706 keys[stat_cnt] = "tx_bytes";
4707 values[stat_cnt] = tx_bytes;
4711 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);