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 "meta-flow.h"
39 #include "ofp-print.h"
42 #include "ofproto/bond.h"
43 #include "ofproto/ofproto.h"
44 #include "poll-loop.h"
49 #include "socket-util.h"
51 #include "stream-ssl.h"
53 #include "system-stats.h"
58 #include "lib/vswitch-idl.h"
59 #include "xenserver.h"
61 #include "sflow_api.h"
62 #include "vlan-bitmap.h"
64 VLOG_DEFINE_THIS_MODULE(bridge);
66 COVERAGE_DEFINE(bridge_reconfigure);
69 /* These members are always valid.
71 * They are immutable: they never change between iface_create() and
73 struct list port_elem; /* Element in struct port's "ifaces" list. */
74 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
75 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
76 struct port *port; /* Containing port. */
77 char *name; /* Host network device name. */
78 struct netdev *netdev; /* Network device. */
79 ofp_port_t ofp_port; /* OpenFlow port number. */
82 /* These members are valid only within bridge_reconfigure(). */
83 const char *type; /* Usually same as cfg->type. */
84 const struct ovsrec_interface *cfg;
88 struct uuid uuid; /* UUID of this "mirror" record in database. */
89 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
90 struct bridge *bridge;
92 const struct ovsrec_mirror *cfg;
96 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
97 struct bridge *bridge;
100 const struct ovsrec_port *cfg;
102 /* An ordinary bridge port has 1 interface.
103 * A bridge port for bonding has at least 2 interfaces. */
104 struct list ifaces; /* List of "struct iface"s. */
108 struct hmap_node node; /* In 'all_bridges'. */
109 char *name; /* User-specified arbitrary name. */
110 char *type; /* Datapath type. */
111 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
112 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
113 const struct ovsrec_bridge *cfg;
115 /* OpenFlow switch processing. */
116 struct ofproto *ofproto; /* OpenFlow switch. */
119 struct hmap ports; /* "struct port"s indexed by name. */
120 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
121 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
123 /* Port mirroring. */
124 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
126 /* Used during reconfiguration. */
127 struct shash wanted_ports;
129 /* Synthetic local port if necessary. */
130 struct ovsrec_port synth_local_port;
131 struct ovsrec_interface synth_local_iface;
132 struct ovsrec_interface *synth_local_ifacep;
135 /* All bridges, indexed by name. */
136 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
138 /* OVSDB IDL used to obtain configuration. */
139 static struct ovsdb_idl *idl;
141 /* We want to complete daemonization, fully detaching from our parent process,
142 * only after we have completed our initial configuration, committed our state
143 * to the database, and received confirmation back from the database server
144 * that it applied the commit. This allows our parent process to know that,
145 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
146 * to have already been filled in. (It is only "very likely" rather than
147 * certain because there is always a slim possibility that the transaction will
148 * fail or that some other client has added new bridges, ports, etc. while
149 * ovs-vswitchd was configuring using an old configuration.)
151 * We only need to do this once for our initial configuration at startup, so
152 * 'initial_config_done' tracks whether we've already done it. While we are
153 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
154 * itself and is otherwise NULL. */
155 static bool initial_config_done;
156 static struct ovsdb_idl_txn *daemonize_txn;
158 /* Most recently processed IDL sequence number. */
159 static unsigned int idl_seqno;
161 /* Track changes to port connectivity. */
162 static uint64_t connectivity_seqno = LLONG_MIN;
164 /* Status update to database.
166 * Some information in the database must be kept as up-to-date as possible to
167 * allow controllers to respond rapidly to network outages. Those status are
168 * updated via the 'status_txn'.
170 * We use the global connectivity sequence number to detect the status change.
171 * Also, to prevent the status update from sending too much to the database,
172 * we check the return status of each update transaction and do not start new
173 * update if the previous transaction status is 'TXN_INCOMPLETE'.
175 * 'statux_txn' is NULL if there is no ongoing status update.
177 static struct ovsdb_idl_txn *status_txn;
179 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
180 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
181 #define STATUS_CHECK_AGAIN_MSEC 100
183 /* Each time this timer expires, the bridge fetches interface and mirror
184 * statistics and pushes them into the database. */
185 static int stats_timer_interval;
186 static long long int stats_timer = LLONG_MIN;
188 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
189 * expensive. This can cause bridge_reconfigure() to take a long time during
190 * which no other work can be done. To deal with this problem, we limit port
191 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
192 * call to bridge_reconfigure(). If there is more work to do after the limit
193 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
194 * This allows the rest of the code to catch up on important things like
195 * forwarding packets. */
196 #define OFP_PORT_ACTION_WINDOW 10
198 static void add_del_bridges(const struct ovsrec_open_vswitch *);
199 static void bridge_run__(void);
200 static void bridge_create(const struct ovsrec_bridge *);
201 static void bridge_destroy(struct bridge *);
202 static struct bridge *bridge_lookup(const char *name);
203 static unixctl_cb_func bridge_unixctl_dump_flows;
204 static unixctl_cb_func bridge_unixctl_reconnect;
205 static size_t bridge_get_controllers(const struct bridge *br,
206 struct ovsrec_controller ***controllersp);
207 static void bridge_collect_wanted_ports(struct bridge *,
208 const unsigned long *splinter_vlans,
209 struct shash *wanted_ports);
210 static void bridge_delete_ofprotos(void);
211 static void bridge_delete_or_reconfigure_ports(struct bridge *);
212 static void bridge_del_ports(struct bridge *,
213 const struct shash *wanted_ports);
214 static void bridge_add_ports(struct bridge *,
215 const struct shash *wanted_ports);
217 static void bridge_configure_datapath_id(struct bridge *);
218 static void bridge_configure_netflow(struct bridge *);
219 static void bridge_configure_forward_bpdu(struct bridge *);
220 static void bridge_configure_mac_table(struct bridge *);
221 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
222 static void bridge_configure_ipfix(struct bridge *);
223 static void bridge_configure_stp(struct bridge *);
224 static void bridge_configure_tables(struct bridge *);
225 static void bridge_configure_dp_desc(struct bridge *);
226 static void bridge_configure_remotes(struct bridge *,
227 const struct sockaddr_in *managers,
229 static void bridge_pick_local_hw_addr(struct bridge *,
230 uint8_t ea[ETH_ADDR_LEN],
231 struct iface **hw_addr_iface);
232 static uint64_t bridge_pick_datapath_id(struct bridge *,
233 const uint8_t bridge_ea[ETH_ADDR_LEN],
234 struct iface *hw_addr_iface);
235 static uint64_t dpid_from_hash(const void *, size_t nbytes);
236 static bool bridge_has_bond_fake_iface(const struct bridge *,
238 static bool port_is_bond_fake_iface(const struct port *);
240 static unixctl_cb_func qos_unixctl_show;
242 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
243 static void port_del_ifaces(struct port *);
244 static void port_destroy(struct port *);
245 static struct port *port_lookup(const struct bridge *, const char *name);
246 static void port_configure(struct port *);
247 static struct lacp_settings *port_configure_lacp(struct port *,
248 struct lacp_settings *);
249 static void port_configure_bond(struct port *, struct bond_settings *);
250 static bool port_is_synthetic(const struct port *);
252 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
253 static void run_system_stats(void);
255 static void bridge_configure_mirrors(struct bridge *);
256 static struct mirror *mirror_create(struct bridge *,
257 const struct ovsrec_mirror *);
258 static void mirror_destroy(struct mirror *);
259 static bool mirror_configure(struct mirror *);
260 static void mirror_refresh_stats(struct mirror *);
262 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
263 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
264 const struct ovsrec_port *);
265 static bool iface_is_internal(const struct ovsrec_interface *iface,
266 const struct ovsrec_bridge *br);
267 static const char *iface_get_type(const struct ovsrec_interface *,
268 const struct ovsrec_bridge *);
269 static void iface_destroy(struct iface *);
270 static void iface_destroy__(struct iface *);
271 static struct iface *iface_lookup(const struct bridge *, const char *name);
272 static struct iface *iface_find(const char *name);
273 static struct iface *iface_from_ofp_port(const struct bridge *,
274 ofp_port_t ofp_port);
275 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
276 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
277 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
278 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
279 static void iface_configure_cfm(struct iface *);
280 static void iface_refresh_cfm_stats(struct iface *);
281 static void iface_refresh_stats(struct iface *);
282 static void iface_refresh_netdev_status(struct iface *);
283 static void iface_refresh_ofproto_status(struct iface *);
284 static bool iface_is_synthetic(const struct iface *);
285 static ofp_port_t iface_get_requested_ofp_port(
286 const struct ovsrec_interface *);
287 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
289 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
291 * This is deprecated. It is only for compatibility with broken device drivers
292 * in old versions of Linux that do not properly support VLANs when VLAN
293 * devices are not used. When broken device drivers are no longer in
294 * widespread use, we will delete these interfaces. */
296 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
297 static bool vlan_splinters_enabled_anywhere;
299 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
300 static unsigned long int *collect_splinter_vlans(
301 const struct ovsrec_open_vswitch *);
302 static void configure_splinter_port(struct port *);
303 static void add_vlan_splinter_ports(struct bridge *,
304 const unsigned long int *splinter_vlans,
305 struct shash *ports);
308 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
310 struct shash iface_hints;
311 static bool initialized = false;
318 shash_init(&iface_hints);
321 for (i = 0; i < cfg->n_bridges; i++) {
322 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
325 for (j = 0; j < br_cfg->n_ports; j++) {
326 struct ovsrec_port *port_cfg = br_cfg->ports[j];
329 for (k = 0; k < port_cfg->n_interfaces; k++) {
330 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
331 struct iface_hint *iface_hint;
333 iface_hint = xmalloc(sizeof *iface_hint);
334 iface_hint->br_name = br_cfg->name;
335 iface_hint->br_type = br_cfg->datapath_type;
336 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
338 shash_add(&iface_hints, if_cfg->name, iface_hint);
344 ofproto_init(&iface_hints);
346 shash_destroy_free_data(&iface_hints);
350 /* Public functions. */
352 /* Initializes the bridge module, configuring it to obtain its configuration
353 * from an OVSDB server accessed over 'remote', which should be a string in a
354 * form acceptable to ovsdb_idl_create(). */
356 bridge_init(const char *remote)
358 /* Create connection to database. */
359 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
360 idl_seqno = ovsdb_idl_get_seqno(idl);
361 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
362 ovsdb_idl_verify_write_only(idl);
364 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
365 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
366 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
367 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
368 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
369 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
370 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
372 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
373 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
374 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
376 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
377 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
378 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
383 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
384 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
385 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
386 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
387 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
388 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
389 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
399 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
401 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
402 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
403 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
404 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
406 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
408 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
410 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
411 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
413 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
414 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
415 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
416 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
418 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
419 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
420 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
421 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
422 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
424 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
426 /* Register unixctl commands. */
427 unixctl_command_register("qos/show", "interface", 1, 1,
428 qos_unixctl_show, NULL);
429 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
430 bridge_unixctl_dump_flows, NULL);
431 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
432 bridge_unixctl_reconnect, NULL);
442 struct bridge *br, *next_br;
444 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
447 ovsdb_idl_destroy(idl);
450 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
451 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
452 * responsible for freeing '*managersp' (with free()).
454 * You may be asking yourself "why does ovs-vswitchd care?", because
455 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
456 * should not be and in fact is not directly involved in that. But
457 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
458 * it has to tell in-band control where the managers are to enable that.
459 * (Thus, only managers connected in-band are collected.)
462 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
463 struct sockaddr_in **managersp, size_t *n_managersp)
465 struct sockaddr_in *managers = NULL;
466 size_t n_managers = 0;
470 /* Collect all of the potential targets from the "targets" columns of the
471 * rows pointed to by "manager_options", excluding any that are
474 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
475 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
477 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
478 sset_find_and_delete(&targets, m->target);
480 sset_add(&targets, m->target);
484 /* Now extract the targets' IP addresses. */
485 if (!sset_is_empty(&targets)) {
488 managers = xmalloc(sset_count(&targets) * sizeof *managers);
489 SSET_FOR_EACH (target, &targets) {
490 struct sockaddr_storage ss;
492 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
494 && ss.ss_family == AF_INET) {
495 managers[n_managers++] = *(struct sockaddr_in *) &ss;
499 sset_destroy(&targets);
501 *managersp = managers;
502 *n_managersp = n_managers;
506 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
508 unsigned long int *splinter_vlans;
509 struct sockaddr_in *managers;
510 struct bridge *br, *next;
511 int sflow_bridge_number;
514 COVERAGE_INC(bridge_reconfigure);
516 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
517 OFPROTO_FLOW_LIMIT_DEFAULT));
518 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
519 OFPROTO_MAX_IDLE_DEFAULT));
522 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
523 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
525 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
526 * to 'ovs_cfg', with only very minimal configuration otherwise.
528 * This is mostly an update to bridge data structures. Nothing is pushed
529 * down to ofproto or lower layers. */
530 add_del_bridges(ovs_cfg);
531 splinter_vlans = collect_splinter_vlans(ovs_cfg);
532 HMAP_FOR_EACH (br, node, &all_bridges) {
533 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
534 bridge_del_ports(br, &br->wanted_ports);
536 free(splinter_vlans);
538 /* Start pushing configuration changes down to the ofproto layer:
540 * - Delete ofprotos that are no longer configured.
542 * - Delete ports that are no longer configured.
544 * - Reconfigure existing ports to their desired configurations, or
545 * delete them if not possible.
547 * We have to do all the deletions before we can do any additions, because
548 * the ports to be added might require resources that will be freed up by
549 * deletions (they might especially overlap in name). */
550 bridge_delete_ofprotos();
551 HMAP_FOR_EACH (br, node, &all_bridges) {
553 bridge_delete_or_reconfigure_ports(br);
557 /* Finish pushing configuration changes to the ofproto layer:
559 * - Create ofprotos that are missing.
561 * - Add ports that are missing. */
562 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
566 error = ofproto_create(br->name, br->type, &br->ofproto);
568 VLOG_ERR("failed to create bridge %s: %s", br->name,
569 ovs_strerror(error));
570 shash_destroy(&br->wanted_ports);
575 HMAP_FOR_EACH (br, node, &all_bridges) {
576 bridge_add_ports(br, &br->wanted_ports);
577 shash_destroy(&br->wanted_ports);
580 reconfigure_system_stats(ovs_cfg);
582 /* Complete the configuration. */
583 sflow_bridge_number = 0;
584 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
585 HMAP_FOR_EACH (br, node, &all_bridges) {
588 /* We need the datapath ID early to allow LACP ports to use it as the
589 * default system ID. */
590 bridge_configure_datapath_id(br);
592 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
595 port_configure(port);
597 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
598 iface_set_ofport(iface->cfg, iface->ofp_port);
599 iface_configure_cfm(iface);
600 iface_configure_qos(iface, port->cfg->qos);
601 iface_set_mac(br, port, iface);
602 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
606 bridge_configure_mirrors(br);
607 bridge_configure_forward_bpdu(br);
608 bridge_configure_mac_table(br);
609 bridge_configure_remotes(br, managers, n_managers);
610 bridge_configure_netflow(br);
611 bridge_configure_sflow(br, &sflow_bridge_number);
612 bridge_configure_ipfix(br);
613 bridge_configure_stp(br);
614 bridge_configure_tables(br);
615 bridge_configure_dp_desc(br);
619 /* The ofproto-dpif provider does some final reconfiguration in its
620 * ->type_run() function. We have to call it before notifying the database
621 * client that reconfiguration is complete, otherwise there is a very
622 * narrow race window in which e.g. ofproto/trace will not recognize the
623 * new configuration (sometimes this causes unit test failures). */
627 /* Delete ofprotos which aren't configured or have the wrong type. Create
628 * ofprotos which don't exist but need to. */
630 bridge_delete_ofprotos(void)
637 /* Delete ofprotos with no bridge or with the wrong type. */
640 ofproto_enumerate_types(&types);
641 SSET_FOR_EACH (type, &types) {
644 ofproto_enumerate_names(type, &names);
645 SSET_FOR_EACH (name, &names) {
646 br = bridge_lookup(name);
647 if (!br || strcmp(type, br->type)) {
648 ofproto_delete(name, type);
652 sset_destroy(&names);
653 sset_destroy(&types);
657 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
659 if (*n >= *allocated) {
660 ports = x2nrealloc(ports, allocated, sizeof *ports);
662 ports[(*n)++] = port;
667 bridge_delete_or_reconfigure_ports(struct bridge *br)
669 struct ofproto_port ofproto_port;
670 struct ofproto_port_dump dump;
672 struct sset ofproto_ports;
673 struct port *port, *port_next;
675 /* List of "ofp_port"s to delete. We make a list instead of deleting them
676 * right away because ofproto implementations aren't necessarily able to
677 * iterate through a changing list of ports in an entirely robust way. */
684 sset_init(&ofproto_ports);
686 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
687 * that are not configured in the database. (This commonly happens when
688 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
690 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
691 * that have the wrong OpenFlow port number (and arrange to add them back
692 * with the correct OpenFlow port number). */
693 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
694 ofp_port_t requested_ofp_port;
697 sset_add(&ofproto_ports, ofproto_port.name);
699 iface = iface_lookup(br, ofproto_port.name);
701 /* No such iface is configured, so we should delete this
704 * As a corner case exception, keep the port if it's a bond fake
706 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
707 && !strcmp(ofproto_port.type, "internal")) {
713 if (strcmp(ofproto_port.type, iface->type)
714 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
715 /* The interface is the wrong type or can't be configured.
720 /* If the requested OpenFlow port for 'iface' changed, and it's not
721 * already the correct port, then we might want to temporarily delete
722 * this interface, so we can add it back again with the new OpenFlow
724 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
725 if (iface->ofp_port != OFPP_LOCAL &&
726 requested_ofp_port != OFPP_NONE &&
727 requested_ofp_port != iface->ofp_port) {
728 ofp_port_t victim_request;
729 struct iface *victim;
731 /* Check for an existing OpenFlow port currently occupying
732 * 'iface''s requested port number. If there isn't one, then
733 * delete this port. Otherwise we need to consider further. */
734 victim = iface_from_ofp_port(br, requested_ofp_port);
739 /* 'victim' is a port currently using 'iface''s requested port
740 * number. Unless 'victim' specifically requested that port
741 * number, too, then we can delete both 'iface' and 'victim'
742 * temporarily. (We'll add both of them back again later with new
743 * OpenFlow port numbers.)
745 * If 'victim' did request port number 'requested_ofp_port', just
746 * like 'iface', then that's a configuration inconsistency that we
747 * can't resolve. We might as well let it keep its current port
749 victim_request = iface_get_requested_ofp_port(victim->cfg);
750 if (victim_request != requested_ofp_port) {
751 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
752 iface_destroy(victim);
761 iface_destroy(iface);
762 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
764 for (i = 0; i < n; i++) {
765 ofproto_port_del(br->ofproto, del[i]);
769 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
770 * that we didn't see when we iterated through the datapath, i.e. ports
771 * that disappeared underneath use. This is an unusual situation, but it
772 * can happen in some cases:
774 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
775 * idea but could happen).
777 * - The port represented a device that disappeared, e.g. a tuntap
778 * device destroyed via "tunctl -d", a physical Ethernet device
779 * whose module was just unloaded via "rmmod", or a virtual NIC for a
780 * VM whose VM was just terminated. */
781 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
782 struct iface *iface, *iface_next;
784 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
785 if (!sset_contains(&ofproto_ports, iface->name)) {
786 iface_destroy__(iface);
790 if (list_is_empty(&port->ifaces)) {
794 sset_destroy(&ofproto_ports);
798 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
799 bool with_requested_port)
801 struct shash_node *port_node;
803 SHASH_FOR_EACH (port_node, wanted_ports) {
804 const struct ovsrec_port *port_cfg = port_node->data;
807 for (i = 0; i < port_cfg->n_interfaces; i++) {
808 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
809 ofp_port_t requested_ofp_port;
811 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
812 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
813 struct iface *iface = iface_lookup(br, iface_cfg->name);
816 iface_create(br, iface_cfg, port_cfg);
824 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
826 /* First add interfaces that request a particular port number. */
827 bridge_add_ports__(br, wanted_ports, true);
829 /* Then add interfaces that want automatic port number assignment.
830 * We add these afterward to avoid accidentally taking a specifically
831 * requested port number. */
832 bridge_add_ports__(br, wanted_ports, false);
836 port_configure(struct port *port)
838 const struct ovsrec_port *cfg = port->cfg;
839 struct bond_settings bond_settings;
840 struct lacp_settings lacp_settings;
841 struct ofproto_bundle_settings s;
844 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
845 configure_splinter_port(port);
854 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
855 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
856 s.slaves[s.n_slaves++] = iface->ofp_port;
861 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
865 /* Get VLAN trunks. */
868 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
872 if (cfg->vlan_mode) {
873 if (!strcmp(cfg->vlan_mode, "access")) {
874 s.vlan_mode = PORT_VLAN_ACCESS;
875 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
876 s.vlan_mode = PORT_VLAN_TRUNK;
877 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
878 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
879 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
880 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
882 /* This "can't happen" because ovsdb-server should prevent it. */
883 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
884 "back to trunk mode", port->name, cfg->vlan_mode);
885 s.vlan_mode = PORT_VLAN_TRUNK;
889 s.vlan_mode = PORT_VLAN_ACCESS;
891 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
895 s.vlan_mode = PORT_VLAN_TRUNK;
898 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
901 /* Get LACP settings. */
902 s.lacp = port_configure_lacp(port, &lacp_settings);
906 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
907 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
908 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
911 s.lacp_slaves = NULL;
914 /* Get bond settings. */
915 if (s.n_slaves > 1) {
916 s.bond = &bond_settings;
917 port_configure_bond(port, &bond_settings);
920 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
921 netdev_set_miimon_interval(iface->netdev, 0);
926 ofproto_bundle_register(port->bridge->ofproto, port, &s);
934 /* Pick local port hardware address and datapath ID for 'br'. */
936 bridge_configure_datapath_id(struct bridge *br)
938 uint8_t ea[ETH_ADDR_LEN];
940 struct iface *local_iface;
941 struct iface *hw_addr_iface;
944 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
945 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
947 int error = netdev_set_etheraddr(local_iface->netdev, ea);
949 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
950 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
951 "Ethernet address: %s",
952 br->name, ovs_strerror(error));
955 memcpy(br->ea, ea, ETH_ADDR_LEN);
957 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
958 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
959 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
960 ofproto_set_datapath_id(br->ofproto, dpid);
963 dpid_string = xasprintf("%016"PRIx64, dpid);
964 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
968 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
971 bridge_get_allowed_versions(struct bridge *br)
973 if (!br->cfg->n_protocols)
976 return ofputil_versions_from_strings(br->cfg->protocols,
977 br->cfg->n_protocols);
980 /* Set NetFlow configuration on 'br'. */
982 bridge_configure_netflow(struct bridge *br)
984 struct ovsrec_netflow *cfg = br->cfg->netflow;
985 struct netflow_options opts;
988 ofproto_set_netflow(br->ofproto, NULL);
992 memset(&opts, 0, sizeof opts);
994 /* Get default NetFlow configuration from datapath.
995 * Apply overrides from 'cfg'. */
996 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
997 if (cfg->engine_type) {
998 opts.engine_type = *cfg->engine_type;
1000 if (cfg->engine_id) {
1001 opts.engine_id = *cfg->engine_id;
1004 /* Configure active timeout interval. */
1005 opts.active_timeout = cfg->active_timeout;
1006 if (!opts.active_timeout) {
1007 opts.active_timeout = -1;
1008 } else if (opts.active_timeout < 0) {
1009 VLOG_WARN("bridge %s: active timeout interval set to negative "
1010 "value, using default instead (%d seconds)", br->name,
1011 NF_ACTIVE_TIMEOUT_DEFAULT);
1012 opts.active_timeout = -1;
1015 /* Add engine ID to interface number to disambiguate bridgs? */
1016 opts.add_id_to_iface = cfg->add_id_to_interface;
1017 if (opts.add_id_to_iface) {
1018 if (opts.engine_id > 0x7f) {
1019 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1020 "another vswitch, choose an engine id less than 128",
1023 if (hmap_count(&br->ports) > 508) {
1024 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1025 "another port when more than 508 ports are used",
1031 sset_init(&opts.collectors);
1032 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1035 if (ofproto_set_netflow(br->ofproto, &opts)) {
1036 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1038 sset_destroy(&opts.collectors);
1041 /* Set sFlow configuration on 'br'. */
1043 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1045 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1046 struct ovsrec_controller **controllers;
1047 struct ofproto_sflow_options oso;
1048 size_t n_controllers;
1052 ofproto_set_sflow(br->ofproto, NULL);
1056 memset(&oso, 0, sizeof oso);
1058 sset_init(&oso.targets);
1059 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1061 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1062 if (cfg->sampling) {
1063 oso.sampling_rate = *cfg->sampling;
1066 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1068 oso.polling_interval = *cfg->polling;
1071 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1073 oso.header_len = *cfg->header;
1076 oso.sub_id = (*sflow_bridge_number)++;
1077 oso.agent_device = cfg->agent;
1079 oso.control_ip = NULL;
1080 n_controllers = bridge_get_controllers(br, &controllers);
1081 for (i = 0; i < n_controllers; i++) {
1082 if (controllers[i]->local_ip) {
1083 oso.control_ip = controllers[i]->local_ip;
1087 ofproto_set_sflow(br->ofproto, &oso);
1089 sset_destroy(&oso.targets);
1092 /* Returns whether a IPFIX row is valid. */
1094 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1096 return ipfix && ipfix->n_targets > 0;
1099 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1101 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1102 const struct bridge *br)
1104 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1107 /* Set IPFIX configuration on 'br'. */
1109 bridge_configure_ipfix(struct bridge *br)
1111 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1112 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1113 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1114 struct ofproto_ipfix_bridge_exporter_options be_opts;
1115 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1116 size_t n_fe_opts = 0;
1118 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1119 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1124 if (!valid_be_cfg && n_fe_opts == 0) {
1125 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1130 memset(&be_opts, 0, sizeof be_opts);
1132 sset_init(&be_opts.targets);
1133 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1135 if (be_cfg->sampling) {
1136 be_opts.sampling_rate = *be_cfg->sampling;
1138 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1140 if (be_cfg->obs_domain_id) {
1141 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1143 if (be_cfg->obs_point_id) {
1144 be_opts.obs_point_id = *be_cfg->obs_point_id;
1146 if (be_cfg->cache_active_timeout) {
1147 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1149 if (be_cfg->cache_max_flows) {
1150 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1154 if (n_fe_opts > 0) {
1155 struct ofproto_ipfix_flow_exporter_options *opts;
1156 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1158 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1159 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1160 opts->collector_set_id = fe_cfg->id;
1161 sset_init(&opts->targets);
1162 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1163 fe_cfg->ipfix->n_targets);
1164 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1165 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1166 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1167 ? *fe_cfg->ipfix->cache_max_flows : 0;
1173 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1177 sset_destroy(&be_opts.targets);
1180 if (n_fe_opts > 0) {
1181 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1183 for (i = 0; i < n_fe_opts; i++) {
1184 sset_destroy(&opts->targets);
1192 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1193 struct ofproto_port_stp_settings *port_s,
1194 int *port_num_counter, unsigned long *port_num_bitmap)
1196 const char *config_str;
1197 struct iface *iface;
1199 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1200 port_s->enable = false;
1203 port_s->enable = true;
1206 /* STP over bonds is not supported. */
1207 if (!list_is_singleton(&port->ifaces)) {
1208 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1210 port_s->enable = false;
1214 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1216 /* Internal ports shouldn't participate in spanning tree, so
1218 if (!strcmp(iface->type, "internal")) {
1219 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1220 port_s->enable = false;
1224 /* STP on mirror output ports is not supported. */
1225 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1226 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1227 port_s->enable = false;
1231 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1233 unsigned long int port_num = strtoul(config_str, NULL, 0);
1234 int port_idx = port_num - 1;
1236 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1237 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1238 port_s->enable = false;
1242 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1243 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1244 port->name, port_num);
1245 port_s->enable = false;
1248 bitmap_set1(port_num_bitmap, port_idx);
1249 port_s->port_num = port_idx;
1251 if (*port_num_counter >= STP_MAX_PORTS) {
1252 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1253 port_s->enable = false;
1257 port_s->port_num = (*port_num_counter)++;
1260 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1262 port_s->path_cost = strtoul(config_str, NULL, 10);
1264 enum netdev_features current;
1267 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1268 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1269 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1272 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1274 port_s->priority = strtoul(config_str, NULL, 0);
1276 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1280 /* Set spanning tree configuration on 'br'. */
1282 bridge_configure_stp(struct bridge *br)
1284 if (!br->cfg->stp_enable) {
1285 ofproto_set_stp(br->ofproto, NULL);
1287 struct ofproto_stp_settings br_s;
1288 const char *config_str;
1290 int port_num_counter;
1291 unsigned long *port_num_bitmap;
1293 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1295 uint8_t ea[ETH_ADDR_LEN];
1297 if (eth_addr_from_string(config_str, ea)) {
1298 br_s.system_id = eth_addr_to_uint64(ea);
1300 br_s.system_id = eth_addr_to_uint64(br->ea);
1301 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1302 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1305 br_s.system_id = eth_addr_to_uint64(br->ea);
1308 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1310 br_s.priority = strtoul(config_str, NULL, 0);
1312 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1315 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1317 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1319 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1322 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1324 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1326 br_s.max_age = STP_DEFAULT_MAX_AGE;
1329 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1331 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1333 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1336 /* Configure STP on the bridge. */
1337 if (ofproto_set_stp(br->ofproto, &br_s)) {
1338 VLOG_ERR("bridge %s: could not enable STP", br->name);
1342 /* Users must either set the port number with the "stp-port-num"
1343 * configuration on all ports or none. If manual configuration
1344 * is not done, then we allocate them sequentially. */
1345 port_num_counter = 0;
1346 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1347 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1348 struct ofproto_port_stp_settings port_s;
1349 struct iface *iface;
1351 port_configure_stp(br->ofproto, port, &port_s,
1352 &port_num_counter, port_num_bitmap);
1354 /* As bonds are not supported, just apply configuration to
1355 * all interfaces. */
1356 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1357 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1359 VLOG_ERR("port %s: could not enable STP", port->name);
1365 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1366 && port_num_counter) {
1367 VLOG_ERR("bridge %s: must manually configure all STP port "
1368 "IDs or none, disabling", br->name);
1369 ofproto_set_stp(br->ofproto, NULL);
1371 bitmap_free(port_num_bitmap);
1376 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1378 const struct port *port = port_lookup(br, name);
1379 return port && port_is_bond_fake_iface(port);
1383 port_is_bond_fake_iface(const struct port *port)
1385 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1389 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1391 struct bridge *br, *next;
1392 struct shash new_br;
1395 /* Collect new bridges' names and types. */
1396 shash_init(&new_br);
1397 for (i = 0; i < cfg->n_bridges; i++) {
1398 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1399 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1401 if (strchr(br_cfg->name, '/')) {
1402 /* Prevent remote ovsdb-server users from accessing arbitrary
1403 * directories, e.g. consider a bridge named "../../../etc/". */
1404 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1406 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1407 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1411 /* Get rid of deleted bridges or those whose types have changed.
1412 * Update 'cfg' of bridges that still exist. */
1413 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1414 br->cfg = shash_find_data(&new_br, br->name);
1415 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1416 br->cfg->datapath_type))) {
1421 /* Add new bridges. */
1422 for (i = 0; i < cfg->n_bridges; i++) {
1423 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1424 struct bridge *br = bridge_lookup(br_cfg->name);
1426 bridge_create(br_cfg);
1430 shash_destroy(&new_br);
1433 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1434 * Returns 0 if successful, otherwise a positive errno value. */
1436 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1437 struct netdev *netdev)
1439 return netdev_set_config(netdev, &iface_cfg->options);
1442 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1443 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1445 * If successful, returns 0 and stores the network device in '*netdevp'. On
1446 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1448 iface_do_create(const struct bridge *br,
1449 const struct ovsrec_interface *iface_cfg,
1450 const struct ovsrec_port *port_cfg,
1451 ofp_port_t *ofp_portp, struct netdev **netdevp)
1453 struct netdev *netdev = NULL;
1456 if (netdev_is_reserved_name(iface_cfg->name)) {
1457 VLOG_WARN("could not create interface %s, name is reserved",
1463 error = netdev_open(iface_cfg->name,
1464 iface_get_type(iface_cfg, br->cfg), &netdev);
1466 VLOG_WARN("could not open network device %s (%s)",
1467 iface_cfg->name, ovs_strerror(error));
1471 error = iface_set_netdev_config(iface_cfg, netdev);
1476 *ofp_portp = iface_pick_ofport(iface_cfg);
1477 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1482 VLOG_INFO("bridge %s: added interface %s on port %d",
1483 br->name, iface_cfg->name, *ofp_portp);
1485 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1486 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1494 netdev_close(netdev);
1498 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1499 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1500 * automatically allocated for the iface. Takes ownership of and
1501 * deallocates 'if_cfg'.
1503 * Return true if an iface is successfully created, false otherwise. */
1505 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1506 const struct ovsrec_port *port_cfg)
1508 struct netdev *netdev;
1509 struct iface *iface;
1510 ofp_port_t ofp_port;
1514 /* Do the bits that can fail up front. */
1515 ovs_assert(!iface_lookup(br, iface_cfg->name));
1516 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1518 iface_clear_db_record(iface_cfg);
1522 /* Get or create the port structure. */
1523 port = port_lookup(br, port_cfg->name);
1525 port = port_create(br, port_cfg);
1528 /* Create the iface structure. */
1529 iface = xzalloc(sizeof *iface);
1530 list_push_back(&port->ifaces, &iface->port_elem);
1531 hmap_insert(&br->iface_by_name, &iface->name_node,
1532 hash_string(iface_cfg->name, 0));
1534 iface->name = xstrdup(iface_cfg->name);
1535 iface->ofp_port = ofp_port;
1536 iface->netdev = netdev;
1537 iface->type = iface_get_type(iface_cfg, br->cfg);
1538 iface->cfg = iface_cfg;
1539 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1540 hash_ofp_port(ofp_port));
1542 /* Populate initial status in database. */
1543 iface_refresh_stats(iface);
1544 iface_refresh_netdev_status(iface);
1546 /* Add bond fake iface if necessary. */
1547 if (port_is_bond_fake_iface(port)) {
1548 struct ofproto_port ofproto_port;
1550 if (ofproto_port_query_by_name(br->ofproto, port->name,
1552 struct netdev *netdev;
1555 error = netdev_open(port->name, "internal", &netdev);
1557 ofp_port_t fake_ofp_port = OFPP_NONE;
1558 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1559 netdev_close(netdev);
1561 VLOG_WARN("could not open network device %s (%s)",
1562 port->name, ovs_strerror(error));
1565 /* Already exists, nothing to do. */
1566 ofproto_port_destroy(&ofproto_port);
1573 /* Set forward BPDU option. */
1575 bridge_configure_forward_bpdu(struct bridge *br)
1577 ofproto_set_forward_bpdu(br->ofproto,
1578 smap_get_bool(&br->cfg->other_config,
1583 /* Set MAC learning table configuration for 'br'. */
1585 bridge_configure_mac_table(struct bridge *br)
1587 const char *idle_time_str;
1590 const char *mac_table_size_str;
1593 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1594 idle_time = (idle_time_str && atoi(idle_time_str)
1595 ? atoi(idle_time_str)
1596 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1598 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1599 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1600 ? atoi(mac_table_size_str)
1603 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1607 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1608 const struct port *fake_br, struct iface **hw_addr_iface)
1610 struct hmapx mirror_output_ports;
1612 bool found_addr = false;
1616 /* Mirror output ports don't participate in picking the local hardware
1617 * address. ofproto can't help us find out whether a given port is a
1618 * mirror output because we haven't configured mirrors yet, so we need to
1619 * accumulate them ourselves. */
1620 hmapx_init(&mirror_output_ports);
1621 for (i = 0; i < br->cfg->n_mirrors; i++) {
1622 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1623 if (m->output_port) {
1624 hmapx_add(&mirror_output_ports, m->output_port);
1628 /* Otherwise choose the minimum non-local MAC address among all of the
1630 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1631 uint8_t iface_ea[ETH_ADDR_LEN];
1632 struct iface *candidate;
1633 struct iface *iface;
1635 /* Mirror output ports don't participate. */
1636 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1640 /* Choose the MAC address to represent the port. */
1642 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1643 /* Find the interface with this Ethernet address (if any) so that
1644 * we can provide the correct devname to the caller. */
1645 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1646 uint8_t candidate_ea[ETH_ADDR_LEN];
1647 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1648 && eth_addr_equals(iface_ea, candidate_ea)) {
1653 /* Choose the interface whose MAC address will represent the port.
1654 * The Linux kernel bonding code always chooses the MAC address of
1655 * the first slave added to a bond, and the Fedora networking
1656 * scripts always add slaves to a bond in alphabetical order, so
1657 * for compatibility we choose the interface with the name that is
1658 * first in alphabetical order. */
1659 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1660 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1665 /* The local port doesn't count (since we're trying to choose its
1666 * MAC address anyway). */
1667 if (iface->ofp_port == OFPP_LOCAL) {
1671 /* For fake bridges we only choose from ports with the same tag */
1672 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1673 if (!port->cfg->tag) {
1676 if (*port->cfg->tag != *fake_br->cfg->tag) {
1682 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1688 /* Compare against our current choice. */
1689 if (!eth_addr_is_multicast(iface_ea) &&
1690 !eth_addr_is_local(iface_ea) &&
1691 !eth_addr_is_reserved(iface_ea) &&
1692 !eth_addr_is_zero(iface_ea) &&
1693 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1695 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1696 *hw_addr_iface = iface;
1702 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1703 *hw_addr_iface = NULL;
1706 hmapx_destroy(&mirror_output_ports);
1710 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1711 struct iface **hw_addr_iface)
1714 *hw_addr_iface = NULL;
1716 /* Did the user request a particular MAC? */
1717 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1718 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1719 if (eth_addr_is_multicast(ea)) {
1720 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1721 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1722 } else if (eth_addr_is_zero(ea)) {
1723 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1729 /* Find a local hw address */
1730 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1733 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1734 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1735 * an interface on 'br', then that interface must be passed in as
1736 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1737 * 'hw_addr_iface' must be passed in as a null pointer. */
1739 bridge_pick_datapath_id(struct bridge *br,
1740 const uint8_t bridge_ea[ETH_ADDR_LEN],
1741 struct iface *hw_addr_iface)
1744 * The procedure for choosing a bridge MAC address will, in the most
1745 * ordinary case, also choose a unique MAC that we can use as a datapath
1746 * ID. In some special cases, though, multiple bridges will end up with
1747 * the same MAC address. This is OK for the bridges, but it will confuse
1748 * the OpenFlow controller, because each datapath needs a unique datapath
1751 * Datapath IDs must be unique. It is also very desirable that they be
1752 * stable from one run to the next, so that policy set on a datapath
1755 const char *datapath_id;
1758 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1759 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1763 if (!hw_addr_iface) {
1765 * A purely internal bridge, that is, one that has no non-virtual
1766 * network devices on it at all, is difficult because it has no
1767 * natural unique identifier at all.
1769 * When the host is a XenServer, we handle this case by hashing the
1770 * host's UUID with the name of the bridge. Names of bridges are
1771 * persistent across XenServer reboots, although they can be reused if
1772 * an internal network is destroyed and then a new one is later
1773 * created, so this is fairly effective.
1775 * When the host is not a XenServer, we punt by using a random MAC
1776 * address on each run.
1778 const char *host_uuid = xenserver_get_host_uuid();
1780 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1781 dpid = dpid_from_hash(combined, strlen(combined));
1787 return eth_addr_to_uint64(bridge_ea);
1791 dpid_from_hash(const void *data, size_t n)
1793 uint8_t hash[SHA1_DIGEST_SIZE];
1795 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1796 sha1_bytes(data, n, hash);
1797 eth_addr_mark_random(hash);
1798 return eth_addr_to_uint64(hash);
1802 iface_refresh_netdev_status(struct iface *iface)
1806 enum netdev_features current;
1807 enum netdev_flags flags;
1808 const char *link_state;
1809 uint8_t mac[ETH_ADDR_LEN];
1810 int64_t bps, mtu_64, ifindex64, link_resets;
1813 if (iface_is_synthetic(iface)) {
1817 if (iface->change_seq == netdev_get_change_seq(iface->netdev)) {
1821 iface->change_seq = netdev_get_change_seq(iface->netdev);
1825 if (!netdev_get_status(iface->netdev, &smap)) {
1826 ovsrec_interface_set_status(iface->cfg, &smap);
1828 ovsrec_interface_set_status(iface->cfg, NULL);
1831 smap_destroy(&smap);
1833 error = netdev_get_flags(iface->netdev, &flags);
1835 const char *state = flags & NETDEV_UP ? "up" : "down";
1837 ovsrec_interface_set_admin_state(iface->cfg, state);
1839 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1842 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1843 ovsrec_interface_set_link_state(iface->cfg, link_state);
1845 link_resets = netdev_get_carrier_resets(iface->netdev);
1846 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1848 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1849 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1851 ovsrec_interface_set_duplex(iface->cfg,
1852 netdev_features_is_full_duplex(current)
1854 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1856 ovsrec_interface_set_duplex(iface->cfg, NULL);
1857 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1860 error = netdev_get_mtu(iface->netdev, &mtu);
1863 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1865 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1868 error = netdev_get_etheraddr(iface->netdev, mac);
1870 char mac_string[32];
1872 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1873 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1875 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1878 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1879 * if there is no valid ifindex number. */
1880 ifindex64 = netdev_get_ifindex(iface->netdev);
1881 if (ifindex64 < 0) {
1884 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1888 iface_refresh_ofproto_status(struct iface *iface)
1893 if (iface_is_synthetic(iface)) {
1897 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1901 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1903 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1906 iface_refresh_cfm_stats(iface);
1909 error = ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1910 iface->ofp_port, &smap);
1912 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1914 smap_destroy(&smap);
1917 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1920 iface_refresh_cfm_stats(struct iface *iface)
1922 const struct ovsrec_interface *cfg = iface->cfg;
1923 struct ofproto_cfm_status status;
1926 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1927 iface->ofp_port, &status);
1929 /* Do nothing if there is no status change since last update. */
1930 } else if (error > 0) {
1931 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1932 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1933 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1934 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1935 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1936 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1938 const char *reasons[CFM_FAULT_N_REASONS];
1939 int64_t cfm_health = status.health;
1940 int64_t cfm_flap_count = status.flap_count;
1941 bool faulted = status.faults != 0;
1944 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1947 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1948 int reason = 1 << i;
1949 if (status.faults & reason) {
1950 reasons[j++] = cfm_fault_reason_to_str(reason);
1953 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1955 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1957 if (status.remote_opstate >= 0) {
1958 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1959 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1961 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1964 ovsrec_interface_set_cfm_remote_mpids(cfg,
1965 (const int64_t *)status.rmps,
1967 if (cfm_health >= 0) {
1968 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1970 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1978 iface_refresh_stats(struct iface *iface)
1980 #define IFACE_STATS \
1981 IFACE_STAT(rx_packets, "rx_packets") \
1982 IFACE_STAT(tx_packets, "tx_packets") \
1983 IFACE_STAT(rx_bytes, "rx_bytes") \
1984 IFACE_STAT(tx_bytes, "tx_bytes") \
1985 IFACE_STAT(rx_dropped, "rx_dropped") \
1986 IFACE_STAT(tx_dropped, "tx_dropped") \
1987 IFACE_STAT(rx_errors, "rx_errors") \
1988 IFACE_STAT(tx_errors, "tx_errors") \
1989 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1990 IFACE_STAT(rx_over_errors, "rx_over_err") \
1991 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1992 IFACE_STAT(collisions, "collisions")
1994 #define IFACE_STAT(MEMBER, NAME) + 1
1995 enum { N_IFACE_STATS = IFACE_STATS };
1997 int64_t values[N_IFACE_STATS];
1998 char *keys[N_IFACE_STATS];
2001 struct netdev_stats stats;
2003 if (iface_is_synthetic(iface)) {
2007 /* Intentionally ignore return value, since errors will set 'stats' to
2008 * all-1s, and we will deal with that correctly below. */
2009 netdev_get_stats(iface->netdev, &stats);
2011 /* Copy statistics into keys[] and values[]. */
2013 #define IFACE_STAT(MEMBER, NAME) \
2014 if (stats.MEMBER != UINT64_MAX) { \
2016 values[n] = stats.MEMBER; \
2021 ovs_assert(n <= N_IFACE_STATS);
2023 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2028 br_refresh_stp_status(struct bridge *br)
2030 struct smap smap = SMAP_INITIALIZER(&smap);
2031 struct ofproto *ofproto = br->ofproto;
2032 struct ofproto_stp_status status;
2034 if (ofproto_get_stp_status(ofproto, &status)) {
2038 if (!status.enabled) {
2039 ovsrec_bridge_set_status(br->cfg, NULL);
2043 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2044 STP_ID_ARGS(status.bridge_id));
2045 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2046 STP_ID_ARGS(status.designated_root));
2047 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2049 ovsrec_bridge_set_status(br->cfg, &smap);
2050 smap_destroy(&smap);
2054 port_refresh_stp_status(struct port *port)
2056 struct ofproto *ofproto = port->bridge->ofproto;
2057 struct iface *iface;
2058 struct ofproto_port_stp_status status;
2061 if (port_is_synthetic(port)) {
2065 /* STP doesn't currently support bonds. */
2066 if (!list_is_singleton(&port->ifaces)) {
2067 ovsrec_port_set_status(port->cfg, NULL);
2071 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2072 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2076 if (!status.enabled) {
2077 ovsrec_port_set_status(port->cfg, NULL);
2081 /* Set Status column. */
2083 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2084 smap_add(&smap, "stp_state", stp_state_name(status.state));
2085 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2086 smap_add(&smap, "stp_role", stp_role_name(status.role));
2087 ovsrec_port_set_status(port->cfg, &smap);
2088 smap_destroy(&smap);
2092 port_refresh_stp_stats(struct port *port)
2094 struct ofproto *ofproto = port->bridge->ofproto;
2095 struct iface *iface;
2096 struct ofproto_port_stp_stats stats;
2098 int64_t int_values[3];
2100 if (port_is_synthetic(port)) {
2104 /* STP doesn't currently support bonds. */
2105 if (!list_is_singleton(&port->ifaces)) {
2109 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2110 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2114 if (!stats.enabled) {
2115 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2119 /* Set Statistics column. */
2120 keys[0] = "stp_tx_count";
2121 int_values[0] = stats.tx_count;
2122 keys[1] = "stp_rx_count";
2123 int_values[1] = stats.rx_count;
2124 keys[2] = "stp_error_count";
2125 int_values[2] = stats.error_count;
2127 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2128 ARRAY_SIZE(int_values));
2132 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2134 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2138 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2140 bool enable = enable_system_stats(cfg);
2142 system_stats_enable(enable);
2144 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2149 run_system_stats(void)
2151 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2154 stats = system_stats_run();
2156 struct ovsdb_idl_txn *txn;
2157 struct ovsdb_datum datum;
2159 txn = ovsdb_idl_txn_create(idl);
2160 ovsdb_datum_from_smap(&datum, stats);
2161 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2163 ovsdb_idl_txn_commit(txn);
2164 ovsdb_idl_txn_destroy(txn);
2171 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2174 case OFPCR12_ROLE_EQUAL:
2176 case OFPCR12_ROLE_MASTER:
2178 case OFPCR12_ROLE_SLAVE:
2180 case OFPCR12_ROLE_NOCHANGE:
2182 return "*** INVALID ROLE ***";
2187 refresh_controller_status(void)
2191 const struct ovsrec_controller *cfg;
2195 /* Accumulate status for controllers on all bridges. */
2196 HMAP_FOR_EACH (br, node, &all_bridges) {
2197 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2200 /* Update each controller in the database with current status. */
2201 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2202 struct ofproto_controller_info *cinfo =
2203 shash_find_data(&info, cfg->target);
2206 struct smap smap = SMAP_INITIALIZER(&smap);
2207 const char **values = cinfo->pairs.values;
2208 const char **keys = cinfo->pairs.keys;
2211 for (i = 0; i < cinfo->pairs.n; i++) {
2212 smap_add(&smap, keys[i], values[i]);
2215 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2216 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2218 ovsrec_controller_set_status(cfg, &smap);
2219 smap_destroy(&smap);
2221 ovsrec_controller_set_is_connected(cfg, false);
2222 ovsrec_controller_set_role(cfg, NULL);
2223 ovsrec_controller_set_status(cfg, NULL);
2227 ofproto_free_ofproto_controller_info(&info);
2237 /* Let each datapath type do the work that it needs to do. */
2239 ofproto_enumerate_types(&types);
2240 SSET_FOR_EACH (type, &types) {
2241 ofproto_type_run(type);
2243 sset_destroy(&types);
2245 /* Let each bridge do the work that it needs to do. */
2246 HMAP_FOR_EACH (br, node, &all_bridges) {
2247 ofproto_run(br->ofproto);
2254 static struct ovsrec_open_vswitch null_cfg;
2255 const struct ovsrec_open_vswitch *cfg;
2257 bool vlan_splinters_changed;
2261 ovsrec_open_vswitch_init(&null_cfg);
2265 if (ovsdb_idl_is_lock_contended(idl)) {
2266 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2267 struct bridge *br, *next_br;
2269 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2270 "disabling this process (pid %ld) until it goes away",
2271 (long int) getpid());
2273 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2276 /* Since we will not be running system_stats_run() in this process
2277 * with the current situation of multiple ovs-vswitchd daemons,
2278 * disable system stats collection. */
2279 system_stats_enable(false);
2281 } else if (!ovsdb_idl_has_lock(idl)) {
2284 cfg = ovsrec_open_vswitch_first(idl);
2286 /* Initialize the ofproto library. This only needs to run once, but
2287 * it must be done after the configuration is set. If the
2288 * initialization has already occurred, bridge_init_ofproto()
2289 * returns immediately. */
2290 bridge_init_ofproto(cfg);
2292 /* Once the value of flow-restore-wait is false, we no longer should
2293 * check its value from the database. */
2294 if (cfg && ofproto_get_flow_restore_wait()) {
2295 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2296 "flow-restore-wait", false));
2301 /* Re-configure SSL. We do this on every trip through the main loop,
2302 * instead of just when the database changes, because the contents of the
2303 * key and certificate files can change without the database changing.
2305 * We do this before bridge_reconfigure() because that function might
2306 * initiate SSL connections and thus requires SSL to be configured. */
2307 if (cfg && cfg->ssl) {
2308 const struct ovsrec_ssl *ssl = cfg->ssl;
2310 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2311 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2314 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2315 * usage has changed. */
2316 vlan_splinters_changed = false;
2317 if (vlan_splinters_enabled_anywhere) {
2318 HMAP_FOR_EACH (br, node, &all_bridges) {
2319 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2320 vlan_splinters_changed = true;
2326 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2327 struct ovsdb_idl_txn *txn;
2329 idl_seqno = ovsdb_idl_get_seqno(idl);
2330 txn = ovsdb_idl_txn_create(idl);
2331 bridge_reconfigure(cfg ? cfg : &null_cfg);
2334 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2337 /* If we are completing our initial configuration for this run
2338 * of ovs-vswitchd, then keep the transaction around to monitor
2339 * it for completion. */
2340 if (initial_config_done) {
2341 ovsdb_idl_txn_commit(txn);
2342 ovsdb_idl_txn_destroy(txn);
2344 initial_config_done = true;
2345 daemonize_txn = txn;
2349 if (daemonize_txn) {
2350 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2351 if (status != TXN_INCOMPLETE) {
2352 ovsdb_idl_txn_destroy(daemonize_txn);
2353 daemonize_txn = NULL;
2355 /* ovs-vswitchd has completed initialization, so allow the
2356 * process that forked us to exit successfully. */
2357 daemonize_complete();
2359 vlog_enable_async();
2361 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2365 /* Statistics update interval should always be greater than or equal to
2368 stats_interval = MAX(smap_get_int(&cfg->other_config,
2369 "stats-update-interval",
2372 stats_interval = 5000;
2374 if (stats_timer_interval != stats_interval) {
2375 stats_timer_interval = stats_interval;
2376 stats_timer = LLONG_MIN;
2379 /* Refresh interface and mirror stats if necessary. */
2380 if (time_msec() >= stats_timer) {
2382 struct ovsdb_idl_txn *txn;
2384 txn = ovsdb_idl_txn_create(idl);
2385 HMAP_FOR_EACH (br, node, &all_bridges) {
2389 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2390 struct iface *iface;
2392 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2393 iface_refresh_stats(iface);
2396 port_refresh_stp_stats(port);
2399 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2400 mirror_refresh_stats(m);
2404 refresh_controller_status();
2405 ovsdb_idl_txn_commit(txn);
2406 ovsdb_idl_txn_destroy(txn); /* XXX */
2409 stats_timer = time_msec() + stats_timer_interval;
2415 /* Check the need to update status. */
2416 seq = seq_read(connectivity_seq_get());
2417 if (seq != connectivity_seqno) {
2418 connectivity_seqno = seq;
2419 status_txn = ovsdb_idl_txn_create(idl);
2420 HMAP_FOR_EACH (br, node, &all_bridges) {
2423 br_refresh_stp_status(br);
2424 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2425 struct iface *iface;
2427 port_refresh_stp_status(port);
2428 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2429 iface_refresh_netdev_status(iface);
2430 iface_refresh_ofproto_status(iface);
2438 enum ovsdb_idl_txn_status status;
2440 status = ovsdb_idl_txn_commit(status_txn);
2441 /* Do not destroy "status_txn" if the transaction is
2442 * "TXN_INCOMPLETE". */
2443 if (status != TXN_INCOMPLETE) {
2444 ovsdb_idl_txn_destroy(status_txn);
2458 ovsdb_idl_wait(idl);
2459 if (daemonize_txn) {
2460 ovsdb_idl_txn_wait(daemonize_txn);
2464 ofproto_enumerate_types(&types);
2465 SSET_FOR_EACH (type, &types) {
2466 ofproto_type_wait(type);
2468 sset_destroy(&types);
2470 if (!hmap_is_empty(&all_bridges)) {
2473 HMAP_FOR_EACH (br, node, &all_bridges) {
2474 ofproto_wait(br->ofproto);
2477 poll_timer_wait_until(stats_timer);
2480 /* If the status database transaction is 'TXN_INCOMPLETE' in this run,
2481 * register a timeout in 'STATUS_CHECK_AGAIN_MSEC'. Else, wait on the
2482 * global connectivity sequence number. Note, this also helps batch
2483 * multiple status changes into one transaction. */
2485 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2487 seq_wait(connectivity_seq_get(), connectivity_seqno);
2490 system_stats_wait();
2493 /* Adds some memory usage statistics for bridges into 'usage', for use with
2494 * memory_report(). */
2496 bridge_get_memory_usage(struct simap *usage)
2503 ofproto_enumerate_types(&types);
2504 SSET_FOR_EACH (type, &types) {
2505 ofproto_type_get_memory_usage(type, usage);
2507 sset_destroy(&types);
2509 HMAP_FOR_EACH (br, node, &all_bridges) {
2510 ofproto_get_memory_usage(br->ofproto, usage);
2514 /* QoS unixctl user interface functions. */
2516 struct qos_unixctl_show_cbdata {
2518 struct iface *iface;
2522 qos_unixctl_show_queue(unsigned int queue_id,
2523 const struct smap *details,
2524 struct iface *iface,
2527 struct netdev_queue_stats stats;
2528 struct smap_node *node;
2531 ds_put_cstr(ds, "\n");
2533 ds_put_format(ds, "Queue %u:\n", queue_id);
2535 ds_put_cstr(ds, "Default:\n");
2538 SMAP_FOR_EACH (node, details) {
2539 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2542 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2544 if (stats.tx_packets != UINT64_MAX) {
2545 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2548 if (stats.tx_bytes != UINT64_MAX) {
2549 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2552 if (stats.tx_errors != UINT64_MAX) {
2553 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2556 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2557 queue_id, ovs_strerror(error));
2562 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2563 const char *argv[], void *aux OVS_UNUSED)
2565 struct ds ds = DS_EMPTY_INITIALIZER;
2566 struct smap smap = SMAP_INITIALIZER(&smap);
2567 struct iface *iface;
2569 struct smap_node *node;
2571 iface = iface_find(argv[1]);
2573 unixctl_command_reply_error(conn, "no such interface");
2577 netdev_get_qos(iface->netdev, &type, &smap);
2579 if (*type != '\0') {
2580 struct netdev_queue_dump dump;
2581 struct smap details;
2582 unsigned int queue_id;
2584 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2586 SMAP_FOR_EACH (node, &smap) {
2587 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2590 smap_init(&details);
2591 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2592 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2594 smap_destroy(&details);
2596 unixctl_command_reply(conn, ds_cstr(&ds));
2598 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2599 unixctl_command_reply_error(conn, ds_cstr(&ds));
2602 smap_destroy(&smap);
2606 /* Bridge reconfiguration functions. */
2608 bridge_create(const struct ovsrec_bridge *br_cfg)
2612 ovs_assert(!bridge_lookup(br_cfg->name));
2613 br = xzalloc(sizeof *br);
2615 br->name = xstrdup(br_cfg->name);
2616 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2619 /* Derive the default Ethernet address from the bridge's UUID. This should
2620 * be unique and it will be stable between ovs-vswitchd runs. */
2621 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2622 eth_addr_mark_random(br->default_ea);
2624 hmap_init(&br->ports);
2625 hmap_init(&br->ifaces);
2626 hmap_init(&br->iface_by_name);
2627 hmap_init(&br->mirrors);
2629 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2633 bridge_destroy(struct bridge *br)
2636 struct mirror *mirror, *next_mirror;
2637 struct port *port, *next_port;
2639 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2642 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2643 mirror_destroy(mirror);
2646 hmap_remove(&all_bridges, &br->node);
2647 ofproto_destroy(br->ofproto);
2648 hmap_destroy(&br->ifaces);
2649 hmap_destroy(&br->ports);
2650 hmap_destroy(&br->iface_by_name);
2651 hmap_destroy(&br->mirrors);
2658 static struct bridge *
2659 bridge_lookup(const char *name)
2663 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2664 if (!strcmp(br->name, name)) {
2671 /* Handle requests for a listing of all flows known by the OpenFlow
2672 * stack, including those normally hidden. */
2674 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2675 const char *argv[], void *aux OVS_UNUSED)
2680 br = bridge_lookup(argv[1]);
2682 unixctl_command_reply_error(conn, "Unknown bridge");
2687 ofproto_get_all_flows(br->ofproto, &results);
2689 unixctl_command_reply(conn, ds_cstr(&results));
2690 ds_destroy(&results);
2693 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2694 * connections and reconnect. If BRIDGE is not specified, then all bridges
2695 * drop their controller connections and reconnect. */
2697 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2698 const char *argv[], void *aux OVS_UNUSED)
2702 br = bridge_lookup(argv[1]);
2704 unixctl_command_reply_error(conn, "Unknown bridge");
2707 ofproto_reconnect_controllers(br->ofproto);
2709 HMAP_FOR_EACH (br, node, &all_bridges) {
2710 ofproto_reconnect_controllers(br->ofproto);
2713 unixctl_command_reply(conn, NULL);
2717 bridge_get_controllers(const struct bridge *br,
2718 struct ovsrec_controller ***controllersp)
2720 struct ovsrec_controller **controllers;
2721 size_t n_controllers;
2723 controllers = br->cfg->controller;
2724 n_controllers = br->cfg->n_controller;
2726 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2732 *controllersp = controllers;
2734 return n_controllers;
2738 bridge_collect_wanted_ports(struct bridge *br,
2739 const unsigned long int *splinter_vlans,
2740 struct shash *wanted_ports)
2744 shash_init(wanted_ports);
2746 for (i = 0; i < br->cfg->n_ports; i++) {
2747 const char *name = br->cfg->ports[i]->name;
2748 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2749 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2753 if (bridge_get_controllers(br, NULL)
2754 && !shash_find(wanted_ports, br->name)) {
2755 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2756 br->name, br->name);
2758 ovsrec_interface_init(&br->synth_local_iface);
2759 ovsrec_port_init(&br->synth_local_port);
2761 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2762 br->synth_local_port.n_interfaces = 1;
2763 br->synth_local_port.name = br->name;
2765 br->synth_local_iface.name = br->name;
2766 br->synth_local_iface.type = "internal";
2768 br->synth_local_ifacep = &br->synth_local_iface;
2770 shash_add(wanted_ports, br->name, &br->synth_local_port);
2773 if (splinter_vlans) {
2774 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2778 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2779 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2780 * 'br' needs to complete its configuration. */
2782 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2784 struct shash_node *port_node;
2785 struct port *port, *next;
2787 /* Get rid of deleted ports.
2788 * Get rid of deleted interfaces on ports that still exist. */
2789 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2790 port->cfg = shash_find_data(wanted_ports, port->name);
2794 port_del_ifaces(port);
2798 /* Update iface->cfg and iface->type in interfaces that still exist. */
2799 SHASH_FOR_EACH (port_node, wanted_ports) {
2800 const struct ovsrec_port *port = port_node->data;
2803 for (i = 0; i < port->n_interfaces; i++) {
2804 const struct ovsrec_interface *cfg = port->interfaces[i];
2805 struct iface *iface = iface_lookup(br, cfg->name);
2806 const char *type = iface_get_type(cfg, br->cfg);
2811 } else if (!strcmp(type, "null")) {
2812 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2813 " may be removed in February 2013. Please email"
2814 " dev@openvswitch.org with concerns.",
2817 /* We will add new interfaces later. */
2823 /* Initializes 'oc' appropriately as a management service controller for
2826 * The caller must free oc->target when it is no longer needed. */
2828 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2829 struct ofproto_controller *oc)
2831 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2832 oc->max_backoff = 0;
2833 oc->probe_interval = 60;
2834 oc->band = OFPROTO_OUT_OF_BAND;
2836 oc->burst_limit = 0;
2837 oc->enable_async_msgs = true;
2840 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2842 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2843 struct ofproto_controller *oc)
2847 oc->target = c->target;
2848 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2849 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2850 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2851 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2852 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2853 oc->burst_limit = (c->controller_burst_limit
2854 ? *c->controller_burst_limit : 0);
2855 oc->enable_async_msgs = (!c->enable_async_messages
2856 || *c->enable_async_messages);
2857 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2858 if (dscp < 0 || dscp > 63) {
2859 dscp = DSCP_DEFAULT;
2864 /* Configures the IP stack for 'br''s local interface properly according to the
2865 * configuration in 'c'. */
2867 bridge_configure_local_iface_netdev(struct bridge *br,
2868 struct ovsrec_controller *c)
2870 struct netdev *netdev;
2871 struct in_addr mask, gateway;
2873 struct iface *local_iface;
2876 /* If there's no local interface or no IP address, give up. */
2877 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2878 if (!local_iface || !c->local_ip
2879 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2883 /* Bring up the local interface. */
2884 netdev = local_iface->netdev;
2885 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2887 /* Configure the IP address and netmask. */
2888 if (!c->local_netmask
2889 || !inet_pton(AF_INET, c->local_netmask, &mask)
2891 mask.s_addr = guess_netmask(ip.s_addr);
2893 if (!netdev_set_in4(netdev, ip, mask)) {
2894 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2895 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2898 /* Configure the default gateway. */
2899 if (c->local_gateway
2900 && inet_pton(AF_INET, c->local_gateway, &gateway)
2901 && gateway.s_addr) {
2902 if (!netdev_add_router(netdev, gateway)) {
2903 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2904 br->name, IP_ARGS(gateway.s_addr));
2909 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2910 * in either string are treated as equal to any number of slashes in the other,
2911 * e.g. "x///y" is equal to "x/y".
2913 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2914 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2915 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2916 * 'b' against a prefix of 'a'.
2919 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2921 const char *b_start = b;
2923 if (b - b_start >= b_stoplen) {
2925 } else if (*a != *b) {
2927 } else if (*a == '/') {
2928 a += strspn(a, "/");
2929 b += strspn(b, "/");
2930 } else if (*a == '\0') {
2940 bridge_configure_remotes(struct bridge *br,
2941 const struct sockaddr_in *managers, size_t n_managers)
2943 bool disable_in_band;
2945 struct ovsrec_controller **controllers;
2946 size_t n_controllers;
2948 enum ofproto_fail_mode fail_mode;
2950 struct ofproto_controller *ocs;
2954 /* Check if we should disable in-band control on this bridge. */
2955 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2958 /* Set OpenFlow queue ID for in-band control. */
2959 ofproto_set_in_band_queue(br->ofproto,
2960 smap_get_int(&br->cfg->other_config,
2961 "in-band-queue", -1));
2963 if (disable_in_band) {
2964 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2966 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2969 n_controllers = bridge_get_controllers(br, &controllers);
2971 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2974 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2975 for (i = 0; i < n_controllers; i++) {
2976 struct ovsrec_controller *c = controllers[i];
2978 if (!strncmp(c->target, "punix:", 6)
2979 || !strncmp(c->target, "unix:", 5)) {
2980 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2983 if (!strncmp(c->target, "unix:", 5)) {
2984 /* Connect to a listening socket */
2985 whitelist = xasprintf("unix:%s/", ovs_rundir());
2986 if (strchr(c->target, '/') &&
2987 !equal_pathnames(c->target, whitelist,
2988 strlen(whitelist))) {
2989 /* Absolute path specified, but not in ovs_rundir */
2990 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
2991 "controller \"%s\" due to possibility for "
2992 "remote exploit. Instead, specify socket "
2993 "in whitelisted \"%s\" or connect to "
2994 "\"unix:%s/%s.mgmt\" (which is always "
2995 "available without special configuration).",
2996 br->name, c->target, whitelist,
2997 ovs_rundir(), br->name);
3002 whitelist = xasprintf("punix:%s/%s.controller",
3003 ovs_rundir(), br->name);
3004 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3005 /* Prevent remote ovsdb-server users from accessing
3006 * arbitrary Unix domain sockets and overwriting arbitrary
3008 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3009 "controller \"%s\" due to possibility of "
3010 "overwriting local files. Instead, specify "
3011 "whitelisted \"%s\" or connect to "
3012 "\"unix:%s/%s.mgmt\" (which is always "
3013 "available without special configuration).",
3014 br->name, c->target, whitelist,
3015 ovs_rundir(), br->name);
3024 bridge_configure_local_iface_netdev(br, c);
3025 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3026 if (disable_in_band) {
3027 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3032 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3033 bridge_get_allowed_versions(br));
3034 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3037 /* Set the fail-mode. */
3038 fail_mode = !br->cfg->fail_mode
3039 || !strcmp(br->cfg->fail_mode, "standalone")
3040 ? OFPROTO_FAIL_STANDALONE
3041 : OFPROTO_FAIL_SECURE;
3042 ofproto_set_fail_mode(br->ofproto, fail_mode);
3044 /* Configure OpenFlow controller connection snooping. */
3045 if (!ofproto_has_snoops(br->ofproto)) {
3049 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3050 ovs_rundir(), br->name));
3051 ofproto_set_snoops(br->ofproto, &snoops);
3052 sset_destroy(&snoops);
3057 bridge_configure_tables(struct bridge *br)
3059 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3063 n_tables = ofproto_get_n_tables(br->ofproto);
3065 for (i = 0; i < n_tables; i++) {
3066 struct ofproto_table_settings s;
3069 s.max_flows = UINT_MAX;
3072 s.n_prefix_fields = 0;
3073 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3075 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3076 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3079 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3080 s.max_flows = *cfg->flow_limit;
3082 if (cfg->overflow_policy
3083 && !strcmp(cfg->overflow_policy, "evict")) {
3085 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3086 for (k = 0; k < cfg->n_groups; k++) {
3087 const char *string = cfg->groups[k];
3090 msg = mf_parse_subfield__(&s.groups[k], &string);
3092 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3093 "'groups' (%s)", br->name, i, msg);
3095 } else if (*string) {
3096 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3097 "element '%s' contains trailing garbage",
3098 br->name, i, cfg->groups[k]);
3104 /* Prefix lookup fields. */
3105 s.n_prefix_fields = 0;
3106 for (k = 0; k < cfg->n_prefixes; k++) {
3107 const char *name = cfg->prefixes[k];
3108 const struct mf_field *mf = mf_from_name(name);
3110 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3114 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3115 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3116 "%s", br->name, name);
3119 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3120 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3121 "field not used: %s", br->name, name);
3124 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3126 if (s.n_prefix_fields > 0) {
3128 struct ds ds = DS_EMPTY_INITIALIZER;
3129 for (k = 0; k < s.n_prefix_fields; k++) {
3131 ds_put_char(&ds, ',');
3133 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3135 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3136 br->name, i, ds_cstr(&ds));
3141 ofproto_configure_table(br->ofproto, i, &s);
3145 for (; j < br->cfg->n_flow_tables; j++) {
3146 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3147 "%"PRId64" not supported by this datapath", br->name,
3148 br->cfg->key_flow_tables[j]);
3153 bridge_configure_dp_desc(struct bridge *br)
3155 ofproto_set_dp_desc(br->ofproto,
3156 smap_get(&br->cfg->other_config, "dp-desc"));
3159 /* Port functions. */
3161 static struct port *
3162 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3166 port = xzalloc(sizeof *port);
3168 port->name = xstrdup(cfg->name);
3170 list_init(&port->ifaces);
3172 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3176 /* Deletes interfaces from 'port' that are no longer configured for it. */
3178 port_del_ifaces(struct port *port)
3180 struct iface *iface, *next;
3181 struct sset new_ifaces;
3184 /* Collect list of new interfaces. */
3185 sset_init(&new_ifaces);
3186 for (i = 0; i < port->cfg->n_interfaces; i++) {
3187 const char *name = port->cfg->interfaces[i]->name;
3188 const char *type = port->cfg->interfaces[i]->type;
3189 if (strcmp(type, "null")) {
3190 sset_add(&new_ifaces, name);
3194 /* Get rid of deleted interfaces. */
3195 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3196 if (!sset_contains(&new_ifaces, iface->name)) {
3197 iface_destroy(iface);
3201 sset_destroy(&new_ifaces);
3205 port_destroy(struct port *port)
3208 struct bridge *br = port->bridge;
3209 struct iface *iface, *next;
3212 ofproto_bundle_unregister(br->ofproto, port);
3215 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3216 iface_destroy__(iface);
3219 hmap_remove(&br->ports, &port->hmap_node);
3225 static struct port *
3226 port_lookup(const struct bridge *br, const char *name)
3230 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3232 if (!strcmp(port->name, name)) {
3240 enable_lacp(struct port *port, bool *activep)
3242 if (!port->cfg->lacp) {
3243 /* XXX when LACP implementation has been sufficiently tested, enable by
3244 * default and make active on bonded ports. */
3246 } else if (!strcmp(port->cfg->lacp, "off")) {
3248 } else if (!strcmp(port->cfg->lacp, "active")) {
3251 } else if (!strcmp(port->cfg->lacp, "passive")) {
3255 VLOG_WARN("port %s: unknown LACP mode %s",
3256 port->name, port->cfg->lacp);
3261 static struct lacp_settings *
3262 port_configure_lacp(struct port *port, struct lacp_settings *s)
3264 const char *lacp_time, *system_id;
3267 if (!enable_lacp(port, &s->active)) {
3271 s->name = port->name;
3273 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3275 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3276 ETH_ADDR_SCAN_ARGS(s->id))) {
3277 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3278 " address.", port->name, system_id);
3282 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3285 if (eth_addr_is_zero(s->id)) {
3286 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3290 /* Prefer bondable links if unspecified. */
3291 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3293 s->priority = (priority > 0 && priority <= UINT16_MAX
3295 : UINT16_MAX - !list_is_short(&port->ifaces));
3297 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3298 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3300 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3301 "lacp-fallback-ab", false);
3307 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3309 int priority, portid, key;
3311 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3312 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3314 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3316 if (portid <= 0 || portid > UINT16_MAX) {
3317 portid = ofp_to_u16(iface->ofp_port);
3320 if (priority <= 0 || priority > UINT16_MAX) {
3321 priority = UINT16_MAX;
3324 if (key < 0 || key > UINT16_MAX) {
3328 s->name = iface->name;
3330 s->priority = priority;
3335 port_configure_bond(struct port *port, struct bond_settings *s)
3337 const char *detect_s;
3338 struct iface *iface;
3339 int miimon_interval;
3341 s->name = port->name;
3343 if (port->cfg->bond_mode) {
3344 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3345 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3346 port->name, port->cfg->bond_mode,
3347 bond_mode_to_string(s->balance));
3350 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3352 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3353 * active-backup. At some point we should remove this warning. */
3354 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3355 " in previous versions, the default bond_mode was"
3356 " balance-slb", port->name,
3357 bond_mode_to_string(s->balance));
3359 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3360 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3361 "please use another bond type or disable flood_vlans",
3365 miimon_interval = smap_get_int(&port->cfg->other_config,
3366 "bond-miimon-interval", 0);
3367 if (miimon_interval <= 0) {
3368 miimon_interval = 200;
3371 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3372 if (!detect_s || !strcmp(detect_s, "carrier")) {
3373 miimon_interval = 0;
3374 } else if (strcmp(detect_s, "miimon")) {
3375 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3376 "defaulting to carrier", port->name, detect_s);
3377 miimon_interval = 0;
3380 s->up_delay = MAX(0, port->cfg->bond_updelay);
3381 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3382 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3383 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3384 "bond-rebalance-interval", 10000);
3385 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3386 s->rebalance_interval = 1000;
3389 s->fake_iface = port->cfg->bond_fake_iface;
3391 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3392 "lacp-fallback-ab", false);
3394 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3395 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3399 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3400 * instead of obtaining it from the database. */
3402 port_is_synthetic(const struct port *port)
3404 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3407 /* Interface functions. */
3410 iface_is_internal(const struct ovsrec_interface *iface,
3411 const struct ovsrec_bridge *br)
3413 /* The local port and "internal" ports are always "internal". */
3414 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3417 /* Returns the correct network device type for interface 'iface' in bridge
3420 iface_get_type(const struct ovsrec_interface *iface,
3421 const struct ovsrec_bridge *br)
3425 /* The local port always has type "internal". Other ports take
3426 * their type from the database and default to "system" if none is
3428 if (iface_is_internal(iface, br)) {
3431 type = iface->type[0] ? iface->type : "system";
3434 return ofproto_port_open_type(br->datapath_type, type);
3438 iface_destroy__(struct iface *iface)
3441 struct port *port = iface->port;
3442 struct bridge *br = port->bridge;
3444 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3445 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3448 if (iface->ofp_port != OFPP_NONE) {
3449 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3452 list_remove(&iface->port_elem);
3453 hmap_remove(&br->iface_by_name, &iface->name_node);
3455 /* The user is changing configuration here, so netdev_remove needs to be
3456 * used as opposed to netdev_close */
3457 netdev_remove(iface->netdev);
3465 iface_destroy(struct iface *iface)
3468 struct port *port = iface->port;
3470 iface_destroy__(iface);
3471 if (list_is_empty(&port->ifaces)) {
3477 static struct iface *
3478 iface_lookup(const struct bridge *br, const char *name)
3480 struct iface *iface;
3482 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3483 &br->iface_by_name) {
3484 if (!strcmp(iface->name, name)) {
3492 static struct iface *
3493 iface_find(const char *name)
3495 const struct bridge *br;
3497 HMAP_FOR_EACH (br, node, &all_bridges) {
3498 struct iface *iface = iface_lookup(br, name);
3507 static struct iface *
3508 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3510 struct iface *iface;
3512 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3514 if (iface->ofp_port == ofp_port) {
3521 /* Set Ethernet address of 'iface', if one is specified in the configuration
3524 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3526 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3527 struct iface *hw_addr_iface;
3529 if (strcmp(iface->type, "internal")) {
3533 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3535 } else if (port->cfg->fake_bridge) {
3536 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3537 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3542 if (iface->ofp_port == OFPP_LOCAL) {
3543 VLOG_ERR("interface %s: ignoring mac in Interface record "
3544 "(use Bridge record to set local port's mac)",
3546 } else if (eth_addr_is_multicast(mac)) {
3547 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3550 int error = netdev_set_etheraddr(iface->netdev, mac);
3552 VLOG_ERR("interface %s: setting MAC failed (%s)",
3553 iface->name, ovs_strerror(error));
3559 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3561 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3563 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3564 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3565 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3569 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3570 * sets the "ofport" field to -1.
3572 * This is appropriate when 'if_cfg''s interface cannot be created or is
3573 * otherwise invalid. */
3575 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3577 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3578 iface_set_ofport(if_cfg, OFPP_NONE);
3579 ovsrec_interface_set_status(if_cfg, NULL);
3580 ovsrec_interface_set_admin_state(if_cfg, NULL);
3581 ovsrec_interface_set_duplex(if_cfg, NULL);
3582 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3583 ovsrec_interface_set_link_state(if_cfg, NULL);
3584 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3585 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3586 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3587 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3588 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3589 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3590 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3591 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3596 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3598 union ovsdb_atom atom;
3600 atom.integer = target;
3601 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3605 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3607 struct ofpbuf queues_buf;
3609 ofpbuf_init(&queues_buf, 0);
3611 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3612 netdev_set_qos(iface->netdev, NULL, NULL);
3614 const struct ovsdb_datum *queues;
3615 struct netdev_queue_dump dump;
3616 unsigned int queue_id;
3617 struct smap details;
3621 /* Configure top-level Qos for 'iface'. */
3622 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3624 /* Deconfigure queues that were deleted. */
3625 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3627 smap_init(&details);
3628 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3629 if (!queue_ids_include(queues, queue_id)) {
3630 netdev_delete_queue(iface->netdev, queue_id);
3633 smap_destroy(&details);
3635 /* Configure queues for 'iface'. */
3637 for (i = 0; i < qos->n_queues; i++) {
3638 const struct ovsrec_queue *queue = qos->value_queues[i];
3639 unsigned int queue_id = qos->key_queues[i];
3641 if (queue_id == 0) {
3645 if (queue->n_dscp == 1) {
3646 struct ofproto_port_queue *port_queue;
3648 port_queue = ofpbuf_put_uninit(&queues_buf,
3649 sizeof *port_queue);
3650 port_queue->queue = queue_id;
3651 port_queue->dscp = queue->dscp[0];
3654 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3657 struct smap details;
3659 smap_init(&details);
3660 netdev_set_queue(iface->netdev, 0, &details);
3661 smap_destroy(&details);
3665 if (iface->ofp_port != OFPP_NONE) {
3666 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3667 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3669 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3670 port_queues, n_queues);
3673 netdev_set_policing(iface->netdev,
3674 iface->cfg->ingress_policing_rate,
3675 iface->cfg->ingress_policing_burst);
3677 ofpbuf_uninit(&queues_buf);
3681 iface_configure_cfm(struct iface *iface)
3683 const struct ovsrec_interface *cfg = iface->cfg;
3684 const char *opstate_str;
3685 const char *cfm_ccm_vlan;
3686 struct cfm_settings s;
3687 struct smap netdev_args;
3689 if (!cfg->n_cfm_mpid) {
3690 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3694 s.check_tnl_key = false;
3695 smap_init(&netdev_args);
3696 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3697 const char *key = smap_get(&netdev_args, "key");
3698 const char *in_key = smap_get(&netdev_args, "in_key");
3700 s.check_tnl_key = (key && !strcmp(key, "flow"))
3701 || (in_key && !strcmp(in_key, "flow"));
3703 smap_destroy(&netdev_args);
3705 s.mpid = *cfg->cfm_mpid;
3706 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3707 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3708 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3710 if (s.interval <= 0) {
3714 if (!cfm_ccm_vlan) {
3716 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3717 s.ccm_vlan = CFM_RANDOM_VLAN;
3719 s.ccm_vlan = atoi(cfm_ccm_vlan);
3720 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3725 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3727 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3729 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3730 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3732 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3735 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3736 * instead of obtaining it from the database. */
3738 iface_is_synthetic(const struct iface *iface)
3740 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3744 iface_validate_ofport__(size_t n, int64_t *ofport)
3746 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3747 ? u16_to_ofp(*ofport)
3752 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3754 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3758 iface_pick_ofport(const struct ovsrec_interface *cfg)
3760 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3761 return (requested_ofport != OFPP_NONE
3763 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3766 /* Port mirroring. */
3768 static struct mirror *
3769 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3773 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3774 if (uuid_equals(uuid, &m->uuid)) {
3782 bridge_configure_mirrors(struct bridge *br)
3784 const struct ovsdb_datum *mc;
3785 unsigned long *flood_vlans;
3786 struct mirror *m, *next;
3789 /* Get rid of deleted mirrors. */
3790 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3791 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3792 union ovsdb_atom atom;
3794 atom.uuid = m->uuid;
3795 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3800 /* Add new mirrors and reconfigure existing ones. */
3801 for (i = 0; i < br->cfg->n_mirrors; i++) {
3802 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3803 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3805 m = mirror_create(br, cfg);
3808 if (!mirror_configure(m)) {
3813 /* Update flooded vlans (for RSPAN). */
3814 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3815 br->cfg->n_flood_vlans);
3816 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3817 bitmap_free(flood_vlans);
3820 static struct mirror *
3821 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3825 m = xzalloc(sizeof *m);
3826 m->uuid = cfg->header_.uuid;
3827 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3829 m->name = xstrdup(cfg->name);
3835 mirror_destroy(struct mirror *m)
3838 struct bridge *br = m->bridge;
3841 ofproto_mirror_unregister(br->ofproto, m);
3844 hmap_remove(&br->mirrors, &m->hmap_node);
3851 mirror_collect_ports(struct mirror *m,
3852 struct ovsrec_port **in_ports, int n_in_ports,
3853 void ***out_portsp, size_t *n_out_portsp)
3855 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3856 size_t n_out_ports = 0;
3859 for (i = 0; i < n_in_ports; i++) {
3860 const char *name = in_ports[i]->name;
3861 struct port *port = port_lookup(m->bridge, name);
3863 out_ports[n_out_ports++] = port;
3865 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3866 "port %s", m->bridge->name, m->name, name);
3869 *out_portsp = out_ports;
3870 *n_out_portsp = n_out_ports;
3874 mirror_configure(struct mirror *m)
3876 const struct ovsrec_mirror *cfg = m->cfg;
3877 struct ofproto_mirror_settings s;
3880 if (strcmp(cfg->name, m->name)) {
3882 m->name = xstrdup(cfg->name);
3886 /* Get output port or VLAN. */
3887 if (cfg->output_port) {
3888 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3889 if (!s.out_bundle) {
3890 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3891 m->bridge->name, m->name);
3894 s.out_vlan = UINT16_MAX;
3896 if (cfg->output_vlan) {
3897 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3898 "output vlan; ignoring output vlan",
3899 m->bridge->name, m->name);
3901 } else if (cfg->output_vlan) {
3902 /* The database should prevent invalid VLAN values. */
3903 s.out_bundle = NULL;
3904 s.out_vlan = *cfg->output_vlan;
3906 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3907 m->bridge->name, m->name);
3911 /* Get port selection. */
3912 if (cfg->select_all) {
3913 size_t n_ports = hmap_count(&m->bridge->ports);
3914 void **ports = xmalloc(n_ports * sizeof *ports);
3919 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3929 /* Get ports, dropping ports that don't exist.
3930 * The IDL ensures that there are no duplicates. */
3931 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3932 &s.srcs, &s.n_srcs);
3933 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3934 &s.dsts, &s.n_dsts);
3937 /* Get VLAN selection. */
3938 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3941 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3944 if (s.srcs != s.dsts) {
3953 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3955 * This is deprecated. It is only for compatibility with broken device drivers
3956 * in old versions of Linux that do not properly support VLANs when VLAN
3957 * devices are not used. When broken device drivers are no longer in
3958 * widespread use, we will delete these interfaces. */
3960 static struct ovsrec_port **recs;
3961 static size_t n_recs, allocated_recs;
3963 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3964 * splinters are reconfigured. */
3966 register_rec(struct ovsrec_port *rec)
3968 if (n_recs >= allocated_recs) {
3969 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3971 recs[n_recs++] = rec;
3974 /* Frees all of the ports registered with register_reg(). */
3976 free_registered_recs(void)
3980 for (i = 0; i < n_recs; i++) {
3981 struct ovsrec_port *port = recs[i];
3984 for (j = 0; j < port->n_interfaces; j++) {
3985 struct ovsrec_interface *iface = port->interfaces[j];
3990 smap_destroy(&port->other_config);
3991 free(port->interfaces);
3999 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4002 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4004 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4008 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4011 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4012 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4013 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4016 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4017 * use, returns NULL.
4019 * Updates 'vlan_splinters_enabled_anywhere'. */
4020 static unsigned long int *
4021 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4023 unsigned long int *splinter_vlans;
4024 struct sset splinter_ifaces;
4025 const char *real_dev_name;
4026 struct shash *real_devs;
4027 struct shash_node *node;
4031 /* Free space allocated for synthesized ports and interfaces, since we're
4032 * in the process of reconstructing all of them. */
4033 free_registered_recs();
4035 splinter_vlans = bitmap_allocate(4096);
4036 sset_init(&splinter_ifaces);
4037 vlan_splinters_enabled_anywhere = false;
4038 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4039 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4042 for (j = 0; j < br_cfg->n_ports; j++) {
4043 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4046 for (k = 0; k < port_cfg->n_interfaces; k++) {
4047 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4049 if (vlan_splinters_is_enabled(iface_cfg)) {
4050 vlan_splinters_enabled_anywhere = true;
4051 sset_add(&splinter_ifaces, iface_cfg->name);
4052 vlan_bitmap_from_array__(port_cfg->trunks,
4058 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4059 bitmap_set1(splinter_vlans, *port_cfg->tag);
4064 if (!vlan_splinters_enabled_anywhere) {
4065 free(splinter_vlans);
4066 sset_destroy(&splinter_ifaces);
4070 HMAP_FOR_EACH (br, node, &all_bridges) {
4072 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4076 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4077 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4078 * device to be created for it. */
4079 bitmap_set0(splinter_vlans, 0);
4080 bitmap_set0(splinter_vlans, 4095);
4082 /* Delete all VLAN devices that we don't need. */
4084 real_devs = vlandev_get_real_devs();
4085 SHASH_FOR_EACH (node, real_devs) {
4086 const struct vlan_real_dev *real_dev = node->data;
4087 const struct vlan_dev *vlan_dev;
4088 bool real_dev_has_splinters;
4090 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4092 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4093 if (!real_dev_has_splinters
4094 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4095 struct netdev *netdev;
4097 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4098 if (!netdev_get_in4(netdev, NULL, NULL) ||
4099 !netdev_get_in6(netdev, NULL)) {
4100 /* It has an IP address configured, so we don't own
4101 * it. Don't delete it. */
4103 vlandev_del(vlan_dev->name);
4105 netdev_close(netdev);
4112 /* Add all VLAN devices that we need. */
4113 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4116 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4117 if (!vlandev_get_name(real_dev_name, vid)) {
4118 vlandev_add(real_dev_name, vid);
4125 sset_destroy(&splinter_ifaces);
4127 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4128 free(splinter_vlans);
4131 return splinter_vlans;
4134 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4137 configure_splinter_port(struct port *port)
4139 struct ofproto *ofproto = port->bridge->ofproto;
4140 ofp_port_t realdev_ofp_port;
4141 const char *realdev_name;
4142 struct iface *vlandev, *realdev;
4144 ofproto_bundle_unregister(port->bridge->ofproto, port);
4146 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4149 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4150 realdev = iface_lookup(port->bridge, realdev_name);
4151 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4153 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4157 static struct ovsrec_port *
4158 synthesize_splinter_port(const char *real_dev_name,
4159 const char *vlan_dev_name, int vid)
4161 struct ovsrec_interface *iface;
4162 struct ovsrec_port *port;
4164 iface = xmalloc(sizeof *iface);
4165 ovsrec_interface_init(iface);
4166 iface->name = xstrdup(vlan_dev_name);
4167 iface->type = "system";
4169 port = xmalloc(sizeof *port);
4170 ovsrec_port_init(port);
4171 port->interfaces = xmemdup(&iface, sizeof iface);
4172 port->n_interfaces = 1;
4173 port->name = xstrdup(vlan_dev_name);
4174 port->vlan_mode = "splinter";
4175 port->tag = xmalloc(sizeof *port->tag);
4178 smap_add(&port->other_config, "realdev", real_dev_name);
4184 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4185 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4186 * 1-bit in the 'splinter_vlans' bitmap. */
4188 add_vlan_splinter_ports(struct bridge *br,
4189 const unsigned long int *splinter_vlans,
4190 struct shash *ports)
4194 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4195 * we're modifying 'ports'. */
4196 for (i = 0; i < br->cfg->n_ports; i++) {
4197 const char *name = br->cfg->ports[i]->name;
4198 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4201 for (j = 0; j < port_cfg->n_interfaces; j++) {
4202 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4204 if (vlan_splinters_is_enabled(iface_cfg)) {
4205 const char *real_dev_name;
4208 real_dev_name = iface_cfg->name;
4209 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4210 const char *vlan_dev_name;
4212 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4214 && !shash_find(ports, vlan_dev_name)) {
4215 shash_add(ports, vlan_dev_name,
4216 synthesize_splinter_port(
4217 real_dev_name, vlan_dev_name, vid));
4226 mirror_refresh_stats(struct mirror *m)
4228 struct ofproto *ofproto = m->bridge->ofproto;
4229 uint64_t tx_packets, tx_bytes;
4232 size_t stat_cnt = 0;
4234 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4235 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4239 if (tx_packets != UINT64_MAX) {
4240 keys[stat_cnt] = "tx_packets";
4241 values[stat_cnt] = tx_packets;
4244 if (tx_bytes != UINT64_MAX) {
4245 keys[stat_cnt] = "tx_bytes";
4246 values[stat_cnt] = tx_bytes;
4250 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);