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. */
81 /* These members are valid only within bridge_reconfigure(). */
82 const char *type; /* Usually same as cfg->type. */
83 const struct ovsrec_interface *cfg;
87 struct uuid uuid; /* UUID of this "mirror" record in database. */
88 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
89 struct bridge *bridge;
91 const struct ovsrec_mirror *cfg;
95 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
96 struct bridge *bridge;
99 const struct ovsrec_port *cfg;
101 /* An ordinary bridge port has 1 interface.
102 * A bridge port for bonding has at least 2 interfaces. */
103 struct list ifaces; /* List of "struct iface"s. */
107 struct hmap_node node; /* In 'all_bridges'. */
108 char *name; /* User-specified arbitrary name. */
109 char *type; /* Datapath type. */
110 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
111 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
112 const struct ovsrec_bridge *cfg;
114 /* OpenFlow switch processing. */
115 struct ofproto *ofproto; /* OpenFlow switch. */
118 struct hmap ports; /* "struct port"s indexed by name. */
119 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
120 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
122 /* Port mirroring. */
123 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
125 /* Used during reconfiguration. */
126 struct shash wanted_ports;
128 /* Synthetic local port if necessary. */
129 struct ovsrec_port synth_local_port;
130 struct ovsrec_interface synth_local_iface;
131 struct ovsrec_interface *synth_local_ifacep;
134 /* All bridges, indexed by name. */
135 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
137 /* OVSDB IDL used to obtain configuration. */
138 static struct ovsdb_idl *idl;
140 /* We want to complete daemonization, fully detaching from our parent process,
141 * only after we have completed our initial configuration, committed our state
142 * to the database, and received confirmation back from the database server
143 * that it applied the commit. This allows our parent process to know that,
144 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
145 * to have already been filled in. (It is only "very likely" rather than
146 * certain because there is always a slim possibility that the transaction will
147 * fail or that some other client has added new bridges, ports, etc. while
148 * ovs-vswitchd was configuring using an old configuration.)
150 * We only need to do this once for our initial configuration at startup, so
151 * 'initial_config_done' tracks whether we've already done it. While we are
152 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
153 * itself and is otherwise NULL. */
154 static bool initial_config_done;
155 static struct ovsdb_idl_txn *daemonize_txn;
157 /* Most recently processed IDL sequence number. */
158 static unsigned int idl_seqno;
160 /* Track changes to port connectivity. */
161 static uint64_t connectivity_seqno = LLONG_MIN;
163 /* Each time this timer expires, the bridge fetches interface and mirror
164 * statistics and pushes them into the database. */
165 #define IFACE_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
166 static long long int iface_stats_timer = LLONG_MIN;
168 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
169 * expensive. This can cause bridge_reconfigure() to take a long time during
170 * which no other work can be done. To deal with this problem, we limit port
171 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
172 * call to bridge_reconfigure(). If there is more work to do after the limit
173 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
174 * This allows the rest of the code to catch up on important things like
175 * forwarding packets. */
176 #define OFP_PORT_ACTION_WINDOW 10
178 static void add_del_bridges(const struct ovsrec_open_vswitch *);
179 static void bridge_run__(void);
180 static void bridge_create(const struct ovsrec_bridge *);
181 static void bridge_destroy(struct bridge *);
182 static struct bridge *bridge_lookup(const char *name);
183 static unixctl_cb_func bridge_unixctl_dump_flows;
184 static unixctl_cb_func bridge_unixctl_reconnect;
185 static size_t bridge_get_controllers(const struct bridge *br,
186 struct ovsrec_controller ***controllersp);
187 static void bridge_collect_wanted_ports(struct bridge *,
188 const unsigned long *splinter_vlans,
189 struct shash *wanted_ports);
190 static void bridge_delete_ofprotos(void);
191 static void bridge_delete_or_reconfigure_ports(struct bridge *);
192 static void bridge_del_ports(struct bridge *,
193 const struct shash *wanted_ports);
194 static void bridge_add_ports(struct bridge *,
195 const struct shash *wanted_ports);
197 static void bridge_configure_flow_miss_model(const char *opt);
198 static void bridge_configure_datapath_id(struct bridge *);
199 static void bridge_configure_netflow(struct bridge *);
200 static void bridge_configure_forward_bpdu(struct bridge *);
201 static void bridge_configure_mac_table(struct bridge *);
202 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
203 static void bridge_configure_ipfix(struct bridge *);
204 static void bridge_configure_stp(struct bridge *);
205 static void bridge_configure_tables(struct bridge *);
206 static void bridge_configure_dp_desc(struct bridge *);
207 static void bridge_configure_remotes(struct bridge *,
208 const struct sockaddr_in *managers,
210 static void bridge_pick_local_hw_addr(struct bridge *,
211 uint8_t ea[ETH_ADDR_LEN],
212 struct iface **hw_addr_iface);
213 static uint64_t bridge_pick_datapath_id(struct bridge *,
214 const uint8_t bridge_ea[ETH_ADDR_LEN],
215 struct iface *hw_addr_iface);
216 static uint64_t dpid_from_hash(const void *, size_t nbytes);
217 static bool bridge_has_bond_fake_iface(const struct bridge *,
219 static bool port_is_bond_fake_iface(const struct port *);
221 static unixctl_cb_func qos_unixctl_show;
223 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
224 static void port_del_ifaces(struct port *);
225 static void port_destroy(struct port *);
226 static struct port *port_lookup(const struct bridge *, const char *name);
227 static void port_configure(struct port *);
228 static struct lacp_settings *port_configure_lacp(struct port *,
229 struct lacp_settings *);
230 static void port_configure_bond(struct port *, struct bond_settings *);
231 static bool port_is_synthetic(const struct port *);
233 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
234 static void run_system_stats(void);
236 static void bridge_configure_mirrors(struct bridge *);
237 static struct mirror *mirror_create(struct bridge *,
238 const struct ovsrec_mirror *);
239 static void mirror_destroy(struct mirror *);
240 static bool mirror_configure(struct mirror *);
241 static void mirror_refresh_stats(struct mirror *);
243 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
244 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
245 const struct ovsrec_port *);
246 static bool iface_is_internal(const struct ovsrec_interface *iface,
247 const struct ovsrec_bridge *br);
248 static const char *iface_get_type(const struct ovsrec_interface *,
249 const struct ovsrec_bridge *);
250 static void iface_destroy(struct iface *);
251 static void iface_destroy__(struct iface *);
252 static struct iface *iface_lookup(const struct bridge *, const char *name);
253 static struct iface *iface_find(const char *name);
254 static struct iface *iface_from_ofp_port(const struct bridge *,
255 ofp_port_t ofp_port);
256 static void iface_set_mac(struct iface *);
257 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
258 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
259 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
260 static void iface_configure_cfm(struct iface *);
261 static void iface_refresh_cfm_stats(struct iface *);
262 static void iface_refresh_stats(struct iface *);
263 static void iface_refresh_status(struct iface *);
264 static bool iface_is_synthetic(const struct iface *);
265 static ofp_port_t iface_get_requested_ofp_port(
266 const struct ovsrec_interface *);
267 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
269 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
271 * This is deprecated. It is only for compatibility with broken device drivers
272 * in old versions of Linux that do not properly support VLANs when VLAN
273 * devices are not used. When broken device drivers are no longer in
274 * widespread use, we will delete these interfaces. */
276 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
277 static bool vlan_splinters_enabled_anywhere;
279 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
280 static unsigned long int *collect_splinter_vlans(
281 const struct ovsrec_open_vswitch *);
282 static void configure_splinter_port(struct port *);
283 static void add_vlan_splinter_ports(struct bridge *,
284 const unsigned long int *splinter_vlans,
285 struct shash *ports);
288 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
290 struct shash iface_hints;
291 static bool initialized = false;
298 shash_init(&iface_hints);
301 for (i = 0; i < cfg->n_bridges; i++) {
302 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
305 for (j = 0; j < br_cfg->n_ports; j++) {
306 struct ovsrec_port *port_cfg = br_cfg->ports[j];
309 for (k = 0; k < port_cfg->n_interfaces; k++) {
310 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
311 struct iface_hint *iface_hint;
313 iface_hint = xmalloc(sizeof *iface_hint);
314 iface_hint->br_name = br_cfg->name;
315 iface_hint->br_type = br_cfg->datapath_type;
316 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
318 shash_add(&iface_hints, if_cfg->name, iface_hint);
324 ofproto_init(&iface_hints);
326 shash_destroy_free_data(&iface_hints);
330 /* Public functions. */
332 /* Initializes the bridge module, configuring it to obtain its configuration
333 * from an OVSDB server accessed over 'remote', which should be a string in a
334 * form acceptable to ovsdb_idl_create(). */
336 bridge_init(const char *remote)
338 /* Create connection to database. */
339 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
340 idl_seqno = ovsdb_idl_get_seqno(idl);
341 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
342 ovsdb_idl_verify_write_only(idl);
344 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
345 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
346 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
347 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
348 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
349 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
352 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
353 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
354 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
356 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
357 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
358 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
359 ovsdb_idl_omit(idl, &ovsrec_port_col_fake_bridge);
361 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
362 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
363 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
380 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
382 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
383 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
384 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
385 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
387 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
389 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
391 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
392 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
394 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
395 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
396 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
397 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
401 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
405 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
407 /* Register unixctl commands. */
408 unixctl_command_register("qos/show", "interface", 1, 1,
409 qos_unixctl_show, NULL);
410 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
411 bridge_unixctl_dump_flows, NULL);
412 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
413 bridge_unixctl_reconnect, NULL);
423 struct bridge *br, *next_br;
425 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
428 ovsdb_idl_destroy(idl);
431 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
432 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
433 * responsible for freeing '*managersp' (with free()).
435 * You may be asking yourself "why does ovs-vswitchd care?", because
436 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
437 * should not be and in fact is not directly involved in that. But
438 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
439 * it has to tell in-band control where the managers are to enable that.
440 * (Thus, only managers connected in-band are collected.)
443 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
444 struct sockaddr_in **managersp, size_t *n_managersp)
446 struct sockaddr_in *managers = NULL;
447 size_t n_managers = 0;
451 /* Collect all of the potential targets from the "targets" columns of the
452 * rows pointed to by "manager_options", excluding any that are
455 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
456 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
458 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
459 sset_find_and_delete(&targets, m->target);
461 sset_add(&targets, m->target);
465 /* Now extract the targets' IP addresses. */
466 if (!sset_is_empty(&targets)) {
469 managers = xmalloc(sset_count(&targets) * sizeof *managers);
470 SSET_FOR_EACH (target, &targets) {
471 struct sockaddr_in *sin = &managers[n_managers];
473 if (stream_parse_target_with_default_port(target,
480 sset_destroy(&targets);
482 *managersp = managers;
483 *n_managersp = n_managers;
487 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
489 unsigned long int *splinter_vlans;
490 struct sockaddr_in *managers;
491 struct bridge *br, *next;
492 int sflow_bridge_number;
495 COVERAGE_INC(bridge_reconfigure);
497 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
498 OFPROTO_FLOW_LIMIT_DEFAULT));
501 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
502 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
504 bridge_configure_flow_miss_model(smap_get(&ovs_cfg->other_config,
505 "force-miss-model"));
507 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
508 * to 'ovs_cfg', with only very minimal configuration otherwise.
510 * This is mostly an update to bridge data structures. Nothing is pushed
511 * down to ofproto or lower layers. */
512 add_del_bridges(ovs_cfg);
513 splinter_vlans = collect_splinter_vlans(ovs_cfg);
514 HMAP_FOR_EACH (br, node, &all_bridges) {
515 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
516 bridge_del_ports(br, &br->wanted_ports);
518 free(splinter_vlans);
520 /* Start pushing configuration changes down to the ofproto layer:
522 * - Delete ofprotos that are no longer configured.
524 * - Delete ports that are no longer configured.
526 * - Reconfigure existing ports to their desired configurations, or
527 * delete them if not possible.
529 * We have to do all the deletions before we can do any additions, because
530 * the ports to be added might require resources that will be freed up by
531 * deletions (they might especially overlap in name). */
532 bridge_delete_ofprotos();
533 HMAP_FOR_EACH (br, node, &all_bridges) {
535 bridge_delete_or_reconfigure_ports(br);
539 /* Finish pushing configuration changes to the ofproto layer:
541 * - Create ofprotos that are missing.
543 * - Add ports that are missing. */
544 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
548 error = ofproto_create(br->name, br->type, &br->ofproto);
550 VLOG_ERR("failed to create bridge %s: %s", br->name,
551 ovs_strerror(error));
552 shash_destroy(&br->wanted_ports);
557 HMAP_FOR_EACH (br, node, &all_bridges) {
558 bridge_add_ports(br, &br->wanted_ports);
559 shash_destroy(&br->wanted_ports);
562 reconfigure_system_stats(ovs_cfg);
564 /* Complete the configuration. */
565 sflow_bridge_number = 0;
566 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
567 HMAP_FOR_EACH (br, node, &all_bridges) {
570 /* We need the datapath ID early to allow LACP ports to use it as the
571 * default system ID. */
572 bridge_configure_datapath_id(br);
574 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
577 port_configure(port);
579 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
580 iface_set_ofport(iface->cfg, iface->ofp_port);
581 iface_configure_cfm(iface);
582 iface_configure_qos(iface, port->cfg->qos);
583 iface_set_mac(iface);
584 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
588 bridge_configure_mirrors(br);
589 bridge_configure_forward_bpdu(br);
590 bridge_configure_mac_table(br);
591 bridge_configure_remotes(br, managers, n_managers);
592 bridge_configure_netflow(br);
593 bridge_configure_sflow(br, &sflow_bridge_number);
594 bridge_configure_ipfix(br);
595 bridge_configure_stp(br);
596 bridge_configure_tables(br);
597 bridge_configure_dp_desc(br);
599 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
600 /* XXX: Remove this warning message eventually. */
601 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
602 " moved to the Open_vSwitch table. Ignoring its"
603 " setting in the bridge table.");
608 /* The ofproto-dpif provider does some final reconfiguration in its
609 * ->type_run() function. We have to call it before notifying the database
610 * client that reconfiguration is complete, otherwise there is a very
611 * narrow race window in which e.g. ofproto/trace will not recognize the
612 * new configuration (sometimes this causes unit test failures). */
616 /* Delete ofprotos which aren't configured or have the wrong type. Create
617 * ofprotos which don't exist but need to. */
619 bridge_delete_ofprotos(void)
626 /* Delete ofprotos with no bridge or with the wrong type. */
629 ofproto_enumerate_types(&types);
630 SSET_FOR_EACH (type, &types) {
633 ofproto_enumerate_names(type, &names);
634 SSET_FOR_EACH (name, &names) {
635 br = bridge_lookup(name);
636 if (!br || strcmp(type, br->type)) {
637 ofproto_delete(name, type);
641 sset_destroy(&names);
642 sset_destroy(&types);
646 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
648 if (*n >= *allocated) {
649 ports = x2nrealloc(ports, allocated, sizeof *ports);
651 ports[(*n)++] = port;
656 bridge_delete_or_reconfigure_ports(struct bridge *br)
658 struct ofproto_port ofproto_port;
659 struct ofproto_port_dump dump;
661 struct sset ofproto_ports;
662 struct port *port, *port_next;
664 /* List of "ofp_port"s to delete. We make a list instead of deleting them
665 * right away because ofproto implementations aren't necessarily able to
666 * iterate through a changing list of ports in an entirely robust way. */
673 sset_init(&ofproto_ports);
675 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
676 * that are not configured in the database. (This commonly happens when
677 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
679 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
680 * that have the wrong OpenFlow port number (and arrange to add them back
681 * with the correct OpenFlow port number). */
682 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
683 ofp_port_t requested_ofp_port;
686 sset_add(&ofproto_ports, ofproto_port.name);
688 iface = iface_lookup(br, ofproto_port.name);
690 /* No such iface is configured, so we should delete this
693 * As a corner case exception, keep the port if it's a bond fake
695 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
696 && !strcmp(ofproto_port.type, "internal")) {
702 if (strcmp(ofproto_port.type, iface->type)
703 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
704 /* The interface is the wrong type or can't be configured.
709 /* If the requested OpenFlow port for 'iface' changed, and it's not
710 * already the correct port, then we might want to temporarily delete
711 * this interface, so we can add it back again with the new OpenFlow
713 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
714 if (iface->ofp_port != OFPP_LOCAL &&
715 requested_ofp_port != OFPP_NONE &&
716 requested_ofp_port != iface->ofp_port) {
717 ofp_port_t victim_request;
718 struct iface *victim;
720 /* Check for an existing OpenFlow port currently occupying
721 * 'iface''s requested port number. If there isn't one, then
722 * delete this port. Otherwise we need to consider further. */
723 victim = iface_from_ofp_port(br, requested_ofp_port);
728 /* 'victim' is a port currently using 'iface''s requested port
729 * number. Unless 'victim' specifically requested that port
730 * number, too, then we can delete both 'iface' and 'victim'
731 * temporarily. (We'll add both of them back again later with new
732 * OpenFlow port numbers.)
734 * If 'victim' did request port number 'requested_ofp_port', just
735 * like 'iface', then that's a configuration inconsistency that we
736 * can't resolve. We might as well let it keep its current port
738 victim_request = iface_get_requested_ofp_port(victim->cfg);
739 if (victim_request != requested_ofp_port) {
740 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
741 iface_destroy(victim);
750 iface_destroy(iface);
751 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
753 for (i = 0; i < n; i++) {
754 ofproto_port_del(br->ofproto, del[i]);
758 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
759 * that we didn't see when we iterated through the datapath, i.e. ports
760 * that disappeared underneath use. This is an unusual situation, but it
761 * can happen in some cases:
763 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
764 * idea but could happen).
766 * - The port represented a device that disappeared, e.g. a tuntap
767 * device destroyed via "tunctl -d", a physical Ethernet device
768 * whose module was just unloaded via "rmmod", or a virtual NIC for a
769 * VM whose VM was just terminated. */
770 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
771 struct iface *iface, *iface_next;
773 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
774 if (!sset_contains(&ofproto_ports, iface->name)) {
775 iface_destroy__(iface);
779 if (list_is_empty(&port->ifaces)) {
783 sset_destroy(&ofproto_ports);
787 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
788 bool with_requested_port)
790 struct shash_node *port_node;
792 SHASH_FOR_EACH (port_node, wanted_ports) {
793 const struct ovsrec_port *port_cfg = port_node->data;
796 for (i = 0; i < port_cfg->n_interfaces; i++) {
797 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
798 ofp_port_t requested_ofp_port;
800 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
801 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
802 struct iface *iface = iface_lookup(br, iface_cfg->name);
805 iface_create(br, iface_cfg, port_cfg);
813 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
815 /* First add interfaces that request a particular port number. */
816 bridge_add_ports__(br, wanted_ports, true);
818 /* Then add interfaces that want automatic port number assignment.
819 * We add these afterward to avoid accidentally taking a specifically
820 * requested port number. */
821 bridge_add_ports__(br, wanted_ports, false);
825 port_configure(struct port *port)
827 const struct ovsrec_port *cfg = port->cfg;
828 struct bond_settings bond_settings;
829 struct lacp_settings lacp_settings;
830 struct ofproto_bundle_settings s;
833 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
834 configure_splinter_port(port);
843 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
844 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
845 s.slaves[s.n_slaves++] = iface->ofp_port;
850 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
854 /* Get VLAN trunks. */
857 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
861 if (cfg->vlan_mode) {
862 if (!strcmp(cfg->vlan_mode, "access")) {
863 s.vlan_mode = PORT_VLAN_ACCESS;
864 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
865 s.vlan_mode = PORT_VLAN_TRUNK;
866 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
867 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
868 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
869 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
871 /* This "can't happen" because ovsdb-server should prevent it. */
872 VLOG_ERR("unknown VLAN mode %s", cfg->vlan_mode);
873 s.vlan_mode = PORT_VLAN_TRUNK;
877 s.vlan_mode = PORT_VLAN_ACCESS;
879 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
883 s.vlan_mode = PORT_VLAN_TRUNK;
886 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
889 /* Get LACP settings. */
890 s.lacp = port_configure_lacp(port, &lacp_settings);
894 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
895 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
896 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
899 s.lacp_slaves = NULL;
902 /* Get bond settings. */
903 if (s.n_slaves > 1) {
904 s.bond = &bond_settings;
905 port_configure_bond(port, &bond_settings);
908 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
909 netdev_set_miimon_interval(iface->netdev, 0);
914 ofproto_bundle_register(port->bridge->ofproto, port, &s);
923 bridge_configure_flow_miss_model(const char *opt)
925 enum ofproto_flow_miss_model model = OFPROTO_HANDLE_MISS_AUTO;
928 if (!strcmp(opt, "with-facets")) {
929 model = OFPROTO_HANDLE_MISS_WITH_FACETS;
930 } else if (!strcmp(opt, "without-facets")) {
931 model = OFPROTO_HANDLE_MISS_WITHOUT_FACETS;
935 ofproto_set_flow_miss_model(model);
938 /* Pick local port hardware address and datapath ID for 'br'. */
940 bridge_configure_datapath_id(struct bridge *br)
942 uint8_t ea[ETH_ADDR_LEN];
944 struct iface *local_iface;
945 struct iface *hw_addr_iface;
948 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
949 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
951 int error = netdev_set_etheraddr(local_iface->netdev, ea);
953 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
954 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
955 "Ethernet address: %s",
956 br->name, ovs_strerror(error));
959 memcpy(br->ea, ea, ETH_ADDR_LEN);
961 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
962 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
963 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
964 ofproto_set_datapath_id(br->ofproto, dpid);
967 dpid_string = xasprintf("%016"PRIx64, dpid);
968 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
972 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
975 bridge_get_allowed_versions(struct bridge *br)
977 if (!br->cfg->n_protocols)
980 return ofputil_versions_from_strings(br->cfg->protocols,
981 br->cfg->n_protocols);
984 /* Set NetFlow configuration on 'br'. */
986 bridge_configure_netflow(struct bridge *br)
988 struct ovsrec_netflow *cfg = br->cfg->netflow;
989 struct netflow_options opts;
992 ofproto_set_netflow(br->ofproto, NULL);
996 memset(&opts, 0, sizeof opts);
998 /* Get default NetFlow configuration from datapath.
999 * Apply overrides from 'cfg'. */
1000 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1001 if (cfg->engine_type) {
1002 opts.engine_type = *cfg->engine_type;
1004 if (cfg->engine_id) {
1005 opts.engine_id = *cfg->engine_id;
1008 /* Configure active timeout interval. */
1009 opts.active_timeout = cfg->active_timeout;
1010 if (!opts.active_timeout) {
1011 opts.active_timeout = -1;
1012 } else if (opts.active_timeout < 0) {
1013 VLOG_WARN("bridge %s: active timeout interval set to negative "
1014 "value, using default instead (%d seconds)", br->name,
1015 NF_ACTIVE_TIMEOUT_DEFAULT);
1016 opts.active_timeout = -1;
1019 /* Add engine ID to interface number to disambiguate bridgs? */
1020 opts.add_id_to_iface = cfg->add_id_to_interface;
1021 if (opts.add_id_to_iface) {
1022 if (opts.engine_id > 0x7f) {
1023 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1024 "another vswitch, choose an engine id less than 128",
1027 if (hmap_count(&br->ports) > 508) {
1028 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1029 "another port when more than 508 ports are used",
1035 sset_init(&opts.collectors);
1036 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1039 if (ofproto_set_netflow(br->ofproto, &opts)) {
1040 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1042 sset_destroy(&opts.collectors);
1045 /* Set sFlow configuration on 'br'. */
1047 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1049 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1050 struct ovsrec_controller **controllers;
1051 struct ofproto_sflow_options oso;
1052 size_t n_controllers;
1056 ofproto_set_sflow(br->ofproto, NULL);
1060 memset(&oso, 0, sizeof oso);
1062 sset_init(&oso.targets);
1063 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1065 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1066 if (cfg->sampling) {
1067 oso.sampling_rate = *cfg->sampling;
1070 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1072 oso.polling_interval = *cfg->polling;
1075 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1077 oso.header_len = *cfg->header;
1080 oso.sub_id = (*sflow_bridge_number)++;
1081 oso.agent_device = cfg->agent;
1083 oso.control_ip = NULL;
1084 n_controllers = bridge_get_controllers(br, &controllers);
1085 for (i = 0; i < n_controllers; i++) {
1086 if (controllers[i]->local_ip) {
1087 oso.control_ip = controllers[i]->local_ip;
1091 ofproto_set_sflow(br->ofproto, &oso);
1093 sset_destroy(&oso.targets);
1096 /* Returns whether a IPFIX row is valid. */
1098 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1100 return ipfix && ipfix->n_targets > 0;
1103 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1105 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1106 const struct bridge *br)
1108 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1111 /* Set IPFIX configuration on 'br'. */
1113 bridge_configure_ipfix(struct bridge *br)
1115 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1116 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1117 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1118 struct ofproto_ipfix_bridge_exporter_options be_opts;
1119 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1120 size_t n_fe_opts = 0;
1122 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1123 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1128 if (!valid_be_cfg && n_fe_opts == 0) {
1129 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1134 memset(&be_opts, 0, sizeof be_opts);
1136 sset_init(&be_opts.targets);
1137 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1139 if (be_cfg->sampling) {
1140 be_opts.sampling_rate = *be_cfg->sampling;
1142 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1144 if (be_cfg->obs_domain_id) {
1145 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1147 if (be_cfg->obs_point_id) {
1148 be_opts.obs_point_id = *be_cfg->obs_point_id;
1150 if (be_cfg->cache_active_timeout) {
1151 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1153 if (be_cfg->cache_max_flows) {
1154 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1158 if (n_fe_opts > 0) {
1159 struct ofproto_ipfix_flow_exporter_options *opts;
1160 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1162 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1163 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1164 opts->collector_set_id = fe_cfg->id;
1165 sset_init(&opts->targets);
1166 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1167 fe_cfg->ipfix->n_targets);
1168 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1169 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1170 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1171 ? *fe_cfg->ipfix->cache_max_flows : 0;
1177 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1181 sset_destroy(&be_opts.targets);
1184 if (n_fe_opts > 0) {
1185 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1187 for (i = 0; i < n_fe_opts; i++) {
1188 sset_destroy(&opts->targets);
1196 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1197 struct ofproto_port_stp_settings *port_s,
1198 int *port_num_counter, unsigned long *port_num_bitmap)
1200 const char *config_str;
1201 struct iface *iface;
1203 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1204 port_s->enable = false;
1207 port_s->enable = true;
1210 /* STP over bonds is not supported. */
1211 if (!list_is_singleton(&port->ifaces)) {
1212 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1214 port_s->enable = false;
1218 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1220 /* Internal ports shouldn't participate in spanning tree, so
1222 if (!strcmp(iface->type, "internal")) {
1223 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1224 port_s->enable = false;
1228 /* STP on mirror output ports is not supported. */
1229 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1230 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1231 port_s->enable = false;
1235 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1237 unsigned long int port_num = strtoul(config_str, NULL, 0);
1238 int port_idx = port_num - 1;
1240 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1241 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1242 port_s->enable = false;
1246 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1247 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1248 port->name, port_num);
1249 port_s->enable = false;
1252 bitmap_set1(port_num_bitmap, port_idx);
1253 port_s->port_num = port_idx;
1255 if (*port_num_counter >= STP_MAX_PORTS) {
1256 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1257 port_s->enable = false;
1261 port_s->port_num = (*port_num_counter)++;
1264 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1266 port_s->path_cost = strtoul(config_str, NULL, 10);
1268 enum netdev_features current;
1271 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1272 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1273 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1276 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1278 port_s->priority = strtoul(config_str, NULL, 0);
1280 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1284 /* Set spanning tree configuration on 'br'. */
1286 bridge_configure_stp(struct bridge *br)
1288 if (!br->cfg->stp_enable) {
1289 ofproto_set_stp(br->ofproto, NULL);
1291 struct ofproto_stp_settings br_s;
1292 const char *config_str;
1294 int port_num_counter;
1295 unsigned long *port_num_bitmap;
1297 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1299 uint8_t ea[ETH_ADDR_LEN];
1301 if (eth_addr_from_string(config_str, ea)) {
1302 br_s.system_id = eth_addr_to_uint64(ea);
1304 br_s.system_id = eth_addr_to_uint64(br->ea);
1305 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1306 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1309 br_s.system_id = eth_addr_to_uint64(br->ea);
1312 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1314 br_s.priority = strtoul(config_str, NULL, 0);
1316 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1319 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1321 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1323 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1326 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1328 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1330 br_s.max_age = STP_DEFAULT_MAX_AGE;
1333 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1335 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1337 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1340 /* Configure STP on the bridge. */
1341 if (ofproto_set_stp(br->ofproto, &br_s)) {
1342 VLOG_ERR("bridge %s: could not enable STP", br->name);
1346 /* Users must either set the port number with the "stp-port-num"
1347 * configuration on all ports or none. If manual configuration
1348 * is not done, then we allocate them sequentially. */
1349 port_num_counter = 0;
1350 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1351 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1352 struct ofproto_port_stp_settings port_s;
1353 struct iface *iface;
1355 port_configure_stp(br->ofproto, port, &port_s,
1356 &port_num_counter, port_num_bitmap);
1358 /* As bonds are not supported, just apply configuration to
1359 * all interfaces. */
1360 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1361 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1363 VLOG_ERR("port %s: could not enable STP", port->name);
1369 if (bitmap_scan(port_num_bitmap, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1370 && port_num_counter) {
1371 VLOG_ERR("bridge %s: must manually configure all STP port "
1372 "IDs or none, disabling", br->name);
1373 ofproto_set_stp(br->ofproto, NULL);
1375 bitmap_free(port_num_bitmap);
1380 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1382 const struct port *port = port_lookup(br, name);
1383 return port && port_is_bond_fake_iface(port);
1387 port_is_bond_fake_iface(const struct port *port)
1389 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1393 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1395 struct bridge *br, *next;
1396 struct shash new_br;
1399 /* Collect new bridges' names and types. */
1400 shash_init(&new_br);
1401 for (i = 0; i < cfg->n_bridges; i++) {
1402 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1403 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1405 if (strchr(br_cfg->name, '/')) {
1406 /* Prevent remote ovsdb-server users from accessing arbitrary
1407 * directories, e.g. consider a bridge named "../../../etc/". */
1408 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1410 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1411 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1415 /* Get rid of deleted bridges or those whose types have changed.
1416 * Update 'cfg' of bridges that still exist. */
1417 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1418 br->cfg = shash_find_data(&new_br, br->name);
1419 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1420 br->cfg->datapath_type))) {
1425 /* Add new bridges. */
1426 for (i = 0; i < cfg->n_bridges; i++) {
1427 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1428 struct bridge *br = bridge_lookup(br_cfg->name);
1430 bridge_create(br_cfg);
1434 shash_destroy(&new_br);
1437 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1438 * Returns 0 if successful, otherwise a positive errno value. */
1440 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1441 struct netdev *netdev)
1443 return netdev_set_config(netdev, &iface_cfg->options);
1446 /* Opens a network device for 'if_cfg' and configures it. If '*ofp_portp'
1447 * is OFPP_NONE, adds the network device to br->ofproto and stores the OpenFlow
1448 * port number in '*ofp_portp'; otherwise leaves br->ofproto and '*ofp_portp'
1451 * If successful, returns 0 and stores the network device in '*netdevp'. On
1452 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1454 iface_do_create(const struct bridge *br,
1455 const struct ovsrec_interface *iface_cfg,
1456 const struct ovsrec_port *port_cfg,
1457 ofp_port_t *ofp_portp, struct netdev **netdevp)
1459 struct netdev *netdev = NULL;
1462 if (netdev_is_reserved_name(iface_cfg->name)) {
1463 VLOG_WARN("could not create interface %s, name is reserved",
1469 error = netdev_open(iface_cfg->name,
1470 iface_get_type(iface_cfg, br->cfg), &netdev);
1472 VLOG_WARN("could not open network device %s (%s)",
1473 iface_cfg->name, ovs_strerror(error));
1477 error = iface_set_netdev_config(iface_cfg, netdev);
1482 *ofp_portp = iface_pick_ofport(iface_cfg);
1483 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1488 VLOG_INFO("bridge %s: added interface %s on port %d",
1489 br->name, iface_cfg->name, *ofp_portp);
1491 if ((port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter"))
1492 || iface_is_internal(iface_cfg, br->cfg)) {
1493 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1501 netdev_close(netdev);
1505 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1506 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1507 * automatically allocated for the iface. Takes ownership of and
1508 * deallocates 'if_cfg'.
1510 * Return true if an iface is successfully created, false otherwise. */
1512 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1513 const struct ovsrec_port *port_cfg)
1515 struct netdev *netdev;
1516 struct iface *iface;
1517 ofp_port_t ofp_port;
1521 /* Do the bits that can fail up front. */
1522 ovs_assert(!iface_lookup(br, iface_cfg->name));
1523 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1525 iface_clear_db_record(iface_cfg);
1529 /* Get or create the port structure. */
1530 port = port_lookup(br, port_cfg->name);
1532 port = port_create(br, port_cfg);
1535 /* Create the iface structure. */
1536 iface = xzalloc(sizeof *iface);
1537 list_push_back(&port->ifaces, &iface->port_elem);
1538 hmap_insert(&br->iface_by_name, &iface->name_node,
1539 hash_string(iface_cfg->name, 0));
1541 iface->name = xstrdup(iface_cfg->name);
1542 iface->ofp_port = ofp_port;
1543 iface->netdev = netdev;
1544 iface->type = iface_get_type(iface_cfg, br->cfg);
1545 iface->cfg = iface_cfg;
1546 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1547 hash_ofp_port(ofp_port));
1549 /* Populate initial status in database. */
1550 iface_refresh_stats(iface);
1551 iface_refresh_status(iface);
1553 /* Add bond fake iface if necessary. */
1554 if (port_is_bond_fake_iface(port)) {
1555 struct ofproto_port ofproto_port;
1557 if (ofproto_port_query_by_name(br->ofproto, port->name,
1559 struct netdev *netdev;
1562 error = netdev_open(port->name, "internal", &netdev);
1564 ofp_port_t fake_ofp_port = OFPP_NONE;
1565 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1566 netdev_close(netdev);
1568 VLOG_WARN("could not open network device %s (%s)",
1569 port->name, ovs_strerror(error));
1572 /* Already exists, nothing to do. */
1573 ofproto_port_destroy(&ofproto_port);
1580 /* Set forward BPDU option. */
1582 bridge_configure_forward_bpdu(struct bridge *br)
1584 ofproto_set_forward_bpdu(br->ofproto,
1585 smap_get_bool(&br->cfg->other_config,
1590 /* Set MAC learning table configuration for 'br'. */
1592 bridge_configure_mac_table(struct bridge *br)
1594 const char *idle_time_str;
1597 const char *mac_table_size_str;
1600 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1601 idle_time = (idle_time_str && atoi(idle_time_str)
1602 ? atoi(idle_time_str)
1603 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1605 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1606 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1607 ? atoi(mac_table_size_str)
1610 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1614 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1615 struct iface **hw_addr_iface)
1617 struct hmapx mirror_output_ports;
1620 bool found_addr = false;
1624 *hw_addr_iface = NULL;
1626 /* Did the user request a particular MAC? */
1627 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1628 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1629 if (eth_addr_is_multicast(ea)) {
1630 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1631 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1632 } else if (eth_addr_is_zero(ea)) {
1633 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1639 /* Mirror output ports don't participate in picking the local hardware
1640 * address. ofproto can't help us find out whether a given port is a
1641 * mirror output because we haven't configured mirrors yet, so we need to
1642 * accumulate them ourselves. */
1643 hmapx_init(&mirror_output_ports);
1644 for (i = 0; i < br->cfg->n_mirrors; i++) {
1645 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1646 if (m->output_port) {
1647 hmapx_add(&mirror_output_ports, m->output_port);
1651 /* Otherwise choose the minimum non-local MAC address among all of the
1653 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1654 uint8_t iface_ea[ETH_ADDR_LEN];
1655 struct iface *candidate;
1656 struct iface *iface;
1658 /* Mirror output ports don't participate. */
1659 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1663 /* Choose the MAC address to represent the port. */
1665 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1666 /* Find the interface with this Ethernet address (if any) so that
1667 * we can provide the correct devname to the caller. */
1668 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1669 uint8_t candidate_ea[ETH_ADDR_LEN];
1670 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1671 && eth_addr_equals(iface_ea, candidate_ea)) {
1676 /* Choose the interface whose MAC address will represent the port.
1677 * The Linux kernel bonding code always chooses the MAC address of
1678 * the first slave added to a bond, and the Fedora networking
1679 * scripts always add slaves to a bond in alphabetical order, so
1680 * for compatibility we choose the interface with the name that is
1681 * first in alphabetical order. */
1682 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1683 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1688 /* The local port doesn't count (since we're trying to choose its
1689 * MAC address anyway). */
1690 if (iface->ofp_port == OFPP_LOCAL) {
1695 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1701 /* Compare against our current choice. */
1702 if (!eth_addr_is_multicast(iface_ea) &&
1703 !eth_addr_is_local(iface_ea) &&
1704 !eth_addr_is_reserved(iface_ea) &&
1705 !eth_addr_is_zero(iface_ea) &&
1706 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1708 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1709 *hw_addr_iface = iface;
1715 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1716 *hw_addr_iface = NULL;
1719 hmapx_destroy(&mirror_output_ports);
1722 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1723 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1724 * an interface on 'br', then that interface must be passed in as
1725 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1726 * 'hw_addr_iface' must be passed in as a null pointer. */
1728 bridge_pick_datapath_id(struct bridge *br,
1729 const uint8_t bridge_ea[ETH_ADDR_LEN],
1730 struct iface *hw_addr_iface)
1733 * The procedure for choosing a bridge MAC address will, in the most
1734 * ordinary case, also choose a unique MAC that we can use as a datapath
1735 * ID. In some special cases, though, multiple bridges will end up with
1736 * the same MAC address. This is OK for the bridges, but it will confuse
1737 * the OpenFlow controller, because each datapath needs a unique datapath
1740 * Datapath IDs must be unique. It is also very desirable that they be
1741 * stable from one run to the next, so that policy set on a datapath
1744 const char *datapath_id;
1747 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1748 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1752 if (!hw_addr_iface) {
1754 * A purely internal bridge, that is, one that has no non-virtual
1755 * network devices on it at all, is difficult because it has no
1756 * natural unique identifier at all.
1758 * When the host is a XenServer, we handle this case by hashing the
1759 * host's UUID with the name of the bridge. Names of bridges are
1760 * persistent across XenServer reboots, although they can be reused if
1761 * an internal network is destroyed and then a new one is later
1762 * created, so this is fairly effective.
1764 * When the host is not a XenServer, we punt by using a random MAC
1765 * address on each run.
1767 const char *host_uuid = xenserver_get_host_uuid();
1769 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1770 dpid = dpid_from_hash(combined, strlen(combined));
1776 return eth_addr_to_uint64(bridge_ea);
1780 dpid_from_hash(const void *data, size_t n)
1782 uint8_t hash[SHA1_DIGEST_SIZE];
1784 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1785 sha1_bytes(data, n, hash);
1786 eth_addr_mark_random(hash);
1787 return eth_addr_to_uint64(hash);
1791 iface_refresh_status(struct iface *iface)
1795 enum netdev_features current;
1799 uint8_t mac[ETH_ADDR_LEN];
1803 if (iface_is_synthetic(iface)) {
1809 if (!netdev_get_status(iface->netdev, &smap)) {
1810 ovsrec_interface_set_status(iface->cfg, &smap);
1812 ovsrec_interface_set_status(iface->cfg, NULL);
1815 smap_destroy(&smap);
1817 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1818 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1820 ovsrec_interface_set_duplex(iface->cfg,
1821 netdev_features_is_full_duplex(current)
1823 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1825 ovsrec_interface_set_duplex(iface->cfg, NULL);
1826 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1829 error = netdev_get_mtu(iface->netdev, &mtu);
1832 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1834 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1837 error = netdev_get_etheraddr(iface->netdev, mac);
1839 char mac_string[32];
1841 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1842 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1844 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1847 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1848 * if there is no valid ifindex number. */
1849 ifindex64 = netdev_get_ifindex(iface->netdev);
1850 if (ifindex64 < 0) {
1853 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1856 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1859 iface_refresh_cfm_stats(struct iface *iface)
1861 const struct ovsrec_interface *cfg = iface->cfg;
1862 struct ofproto_cfm_status status;
1864 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1865 iface->ofp_port, &status)) {
1866 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1867 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1868 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1869 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1870 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1871 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1873 const char *reasons[CFM_FAULT_N_REASONS];
1874 int64_t cfm_health = status.health;
1875 int64_t cfm_flap_count = status.flap_count;
1876 bool faulted = status.faults != 0;
1879 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1882 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1883 int reason = 1 << i;
1884 if (status.faults & reason) {
1885 reasons[j++] = cfm_fault_reason_to_str(reason);
1888 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1890 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1892 if (status.remote_opstate >= 0) {
1893 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1894 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1896 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1899 ovsrec_interface_set_cfm_remote_mpids(cfg,
1900 (const int64_t *)status.rmps,
1902 if (cfm_health >= 0) {
1903 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1905 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1913 iface_refresh_stats(struct iface *iface)
1915 #define IFACE_STATS \
1916 IFACE_STAT(rx_packets, "rx_packets") \
1917 IFACE_STAT(tx_packets, "tx_packets") \
1918 IFACE_STAT(rx_bytes, "rx_bytes") \
1919 IFACE_STAT(tx_bytes, "tx_bytes") \
1920 IFACE_STAT(rx_dropped, "rx_dropped") \
1921 IFACE_STAT(tx_dropped, "tx_dropped") \
1922 IFACE_STAT(rx_errors, "rx_errors") \
1923 IFACE_STAT(tx_errors, "tx_errors") \
1924 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1925 IFACE_STAT(rx_over_errors, "rx_over_err") \
1926 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1927 IFACE_STAT(collisions, "collisions")
1929 #define IFACE_STAT(MEMBER, NAME) + 1
1930 enum { N_IFACE_STATS = IFACE_STATS };
1932 int64_t values[N_IFACE_STATS];
1933 char *keys[N_IFACE_STATS];
1936 struct netdev_stats stats;
1938 if (iface_is_synthetic(iface)) {
1942 /* Intentionally ignore return value, since errors will set 'stats' to
1943 * all-1s, and we will deal with that correctly below. */
1944 netdev_get_stats(iface->netdev, &stats);
1946 /* Copy statistics into keys[] and values[]. */
1948 #define IFACE_STAT(MEMBER, NAME) \
1949 if (stats.MEMBER != UINT64_MAX) { \
1951 values[n] = stats.MEMBER; \
1956 ovs_assert(n <= N_IFACE_STATS);
1958 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1963 br_refresh_stp_status(struct bridge *br)
1965 struct smap smap = SMAP_INITIALIZER(&smap);
1966 struct ofproto *ofproto = br->ofproto;
1967 struct ofproto_stp_status status;
1969 if (ofproto_get_stp_status(ofproto, &status)) {
1973 if (!status.enabled) {
1974 ovsrec_bridge_set_status(br->cfg, NULL);
1978 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1979 STP_ID_ARGS(status.bridge_id));
1980 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1981 STP_ID_ARGS(status.designated_root));
1982 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
1984 ovsrec_bridge_set_status(br->cfg, &smap);
1985 smap_destroy(&smap);
1989 port_refresh_stp_status(struct port *port)
1991 struct ofproto *ofproto = port->bridge->ofproto;
1992 struct iface *iface;
1993 struct ofproto_port_stp_status status;
1996 if (port_is_synthetic(port)) {
2000 /* STP doesn't currently support bonds. */
2001 if (!list_is_singleton(&port->ifaces)) {
2002 ovsrec_port_set_status(port->cfg, NULL);
2006 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2007 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2011 if (!status.enabled) {
2012 ovsrec_port_set_status(port->cfg, NULL);
2016 /* Set Status column. */
2018 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2019 smap_add(&smap, "stp_state", stp_state_name(status.state));
2020 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2021 smap_add(&smap, "stp_role", stp_role_name(status.role));
2022 ovsrec_port_set_status(port->cfg, &smap);
2023 smap_destroy(&smap);
2027 port_refresh_stp_stats(struct port *port)
2029 struct ofproto *ofproto = port->bridge->ofproto;
2030 struct iface *iface;
2031 struct ofproto_port_stp_stats stats;
2033 int64_t int_values[3];
2035 if (port_is_synthetic(port)) {
2039 /* STP doesn't currently support bonds. */
2040 if (!list_is_singleton(&port->ifaces)) {
2044 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2045 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2049 if (!stats.enabled) {
2050 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2054 /* Set Statistics column. */
2055 keys[0] = "stp_tx_count";
2056 int_values[0] = stats.tx_count;
2057 keys[1] = "stp_rx_count";
2058 int_values[1] = stats.rx_count;
2059 keys[2] = "stp_error_count";
2060 int_values[2] = stats.error_count;
2062 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2063 ARRAY_SIZE(int_values));
2067 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2069 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2073 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2075 bool enable = enable_system_stats(cfg);
2077 system_stats_enable(enable);
2079 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2084 run_system_stats(void)
2086 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2089 stats = system_stats_run();
2091 struct ovsdb_idl_txn *txn;
2092 struct ovsdb_datum datum;
2094 txn = ovsdb_idl_txn_create(idl);
2095 ovsdb_datum_from_smap(&datum, stats);
2096 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2098 ovsdb_idl_txn_commit(txn);
2099 ovsdb_idl_txn_destroy(txn);
2106 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2109 case OFPCR12_ROLE_EQUAL:
2111 case OFPCR12_ROLE_MASTER:
2113 case OFPCR12_ROLE_SLAVE:
2115 case OFPCR12_ROLE_NOCHANGE:
2117 return "*** INVALID ROLE ***";
2122 refresh_controller_status(void)
2126 const struct ovsrec_controller *cfg;
2130 /* Accumulate status for controllers on all bridges. */
2131 HMAP_FOR_EACH (br, node, &all_bridges) {
2132 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2135 /* Update each controller in the database with current status. */
2136 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2137 struct ofproto_controller_info *cinfo =
2138 shash_find_data(&info, cfg->target);
2141 struct smap smap = SMAP_INITIALIZER(&smap);
2142 const char **values = cinfo->pairs.values;
2143 const char **keys = cinfo->pairs.keys;
2146 for (i = 0; i < cinfo->pairs.n; i++) {
2147 smap_add(&smap, keys[i], values[i]);
2150 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2151 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2153 ovsrec_controller_set_status(cfg, &smap);
2154 smap_destroy(&smap);
2156 ovsrec_controller_set_is_connected(cfg, false);
2157 ovsrec_controller_set_role(cfg, NULL);
2158 ovsrec_controller_set_status(cfg, NULL);
2162 ofproto_free_ofproto_controller_info(&info);
2167 * Some information in the database must be kept as up-to-date as possible to
2168 * allow controllers to respond rapidly to network outages. We call these
2169 * statistics "instant" stats.
2171 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2172 * assuming that they've changed. The only means we have to determine whether
2173 * they have changed are:
2175 * - Try to commit changes to the database. If nothing changed, then
2176 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2179 * - instant_stats_run() is called late in the run loop, after anything that
2180 * might change any of the instant stats.
2182 * We use these two facts together to avoid waking the process up every
2183 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2186 /* Minimum interval between writing updates to the instant stats to the
2188 #define INSTANT_INTERVAL_MSEC 100
2190 /* Current instant stats database transaction, NULL if there is no ongoing
2192 static struct ovsdb_idl_txn *instant_txn;
2194 /* Next time (in msec on monotonic clock) at which we will update the instant
2196 static long long int instant_next_txn = LLONG_MIN;
2198 /* True if the run loop has run since we last saw that the instant stats were
2199 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2200 * to refresh the instant stats. */
2201 static bool instant_stats_could_have_changed;
2204 instant_stats_run(void)
2206 enum ovsdb_idl_txn_status status;
2208 instant_stats_could_have_changed = true;
2214 if (time_msec() < instant_next_txn) {
2217 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2219 seq = seq_read(connectivity_seq_get());
2220 if (seq == connectivity_seqno) {
2223 connectivity_seqno = seq;
2225 instant_txn = ovsdb_idl_txn_create(idl);
2226 HMAP_FOR_EACH (br, node, &all_bridges) {
2227 struct iface *iface;
2230 br_refresh_stp_status(br);
2232 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2233 port_refresh_stp_status(port);
2236 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2237 enum netdev_flags flags;
2239 const char *link_state;
2240 int64_t link_resets;
2243 if (iface_is_synthetic(iface)) {
2247 current = ofproto_port_is_lacp_current(br->ofproto,
2251 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2253 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2256 error = netdev_get_flags(iface->netdev, &flags);
2258 const char *state = flags & NETDEV_UP ? "up" : "down";
2259 ovsrec_interface_set_admin_state(iface->cfg, state);
2261 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2264 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2265 ovsrec_interface_set_link_state(iface->cfg, link_state);
2267 link_resets = netdev_get_carrier_resets(iface->netdev);
2268 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2270 iface_refresh_cfm_stats(iface);
2273 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2275 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2276 smap_destroy(&smap);
2281 status = ovsdb_idl_txn_commit(instant_txn);
2282 if (status != TXN_INCOMPLETE) {
2283 ovsdb_idl_txn_destroy(instant_txn);
2286 if (status == TXN_UNCHANGED) {
2287 instant_stats_could_have_changed = false;
2292 instant_stats_wait(void)
2295 ovsdb_idl_txn_wait(instant_txn);
2296 } else if (instant_stats_could_have_changed) {
2297 poll_timer_wait_until(instant_next_txn);
2308 /* Let each datapath type do the work that it needs to do. */
2310 ofproto_enumerate_types(&types);
2311 SSET_FOR_EACH (type, &types) {
2312 ofproto_type_run(type);
2314 sset_destroy(&types);
2316 /* Let each bridge do the work that it needs to do. */
2317 HMAP_FOR_EACH (br, node, &all_bridges) {
2318 ofproto_run(br->ofproto);
2325 static struct ovsrec_open_vswitch null_cfg;
2326 const struct ovsrec_open_vswitch *cfg;
2328 bool vlan_splinters_changed;
2331 ovsrec_open_vswitch_init(&null_cfg);
2335 if (ovsdb_idl_is_lock_contended(idl)) {
2336 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2337 struct bridge *br, *next_br;
2339 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2340 "disabling this process (pid %ld) until it goes away",
2341 (long int) getpid());
2343 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2346 /* Since we will not be running system_stats_run() in this process
2347 * with the current situation of multiple ovs-vswitchd daemons,
2348 * disable system stats collection. */
2349 system_stats_enable(false);
2351 } else if (!ovsdb_idl_has_lock(idl)) {
2354 cfg = ovsrec_open_vswitch_first(idl);
2356 /* Initialize the ofproto library. This only needs to run once, but
2357 * it must be done after the configuration is set. If the
2358 * initialization has already occurred, bridge_init_ofproto()
2359 * returns immediately. */
2360 bridge_init_ofproto(cfg);
2362 /* Once the value of flow-restore-wait is false, we no longer should
2363 * check its value from the database. */
2364 if (cfg && ofproto_get_flow_restore_wait()) {
2365 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2366 "flow-restore-wait", false));
2371 /* Re-configure SSL. We do this on every trip through the main loop,
2372 * instead of just when the database changes, because the contents of the
2373 * key and certificate files can change without the database changing.
2375 * We do this before bridge_reconfigure() because that function might
2376 * initiate SSL connections and thus requires SSL to be configured. */
2377 if (cfg && cfg->ssl) {
2378 const struct ovsrec_ssl *ssl = cfg->ssl;
2380 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2381 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2384 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2385 * usage has changed. */
2386 vlan_splinters_changed = false;
2387 if (vlan_splinters_enabled_anywhere) {
2388 HMAP_FOR_EACH (br, node, &all_bridges) {
2389 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2390 vlan_splinters_changed = true;
2396 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2397 struct ovsdb_idl_txn *txn;
2399 idl_seqno = ovsdb_idl_get_seqno(idl);
2400 txn = ovsdb_idl_txn_create(idl);
2401 bridge_reconfigure(cfg ? cfg : &null_cfg);
2404 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2407 /* If we are completing our initial configuration for this run
2408 * of ovs-vswitchd, then keep the transaction around to monitor
2409 * it for completion. */
2410 if (initial_config_done) {
2411 ovsdb_idl_txn_commit(txn);
2412 ovsdb_idl_txn_destroy(txn);
2414 initial_config_done = true;
2415 daemonize_txn = txn;
2419 if (daemonize_txn) {
2420 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2421 if (status != TXN_INCOMPLETE) {
2422 ovsdb_idl_txn_destroy(daemonize_txn);
2423 daemonize_txn = NULL;
2425 /* ovs-vswitchd has completed initialization, so allow the
2426 * process that forked us to exit successfully. */
2427 daemonize_complete();
2429 vlog_enable_async();
2431 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2435 /* Refresh interface and mirror stats if necessary. */
2436 if (time_msec() >= iface_stats_timer) {
2438 struct ovsdb_idl_txn *txn;
2440 txn = ovsdb_idl_txn_create(idl);
2441 HMAP_FOR_EACH (br, node, &all_bridges) {
2445 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2446 struct iface *iface;
2448 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2449 iface_refresh_stats(iface);
2450 iface_refresh_status(iface);
2453 port_refresh_stp_stats(port);
2456 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2457 mirror_refresh_stats(m);
2461 refresh_controller_status();
2462 ovsdb_idl_txn_commit(txn);
2463 ovsdb_idl_txn_destroy(txn); /* XXX */
2466 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2470 instant_stats_run();
2479 ovsdb_idl_wait(idl);
2480 if (daemonize_txn) {
2481 ovsdb_idl_txn_wait(daemonize_txn);
2485 ofproto_enumerate_types(&types);
2486 SSET_FOR_EACH (type, &types) {
2487 ofproto_type_wait(type);
2489 sset_destroy(&types);
2491 if (!hmap_is_empty(&all_bridges)) {
2494 HMAP_FOR_EACH (br, node, &all_bridges) {
2495 ofproto_wait(br->ofproto);
2497 poll_timer_wait_until(iface_stats_timer);
2500 system_stats_wait();
2501 instant_stats_wait();
2504 /* Adds some memory usage statistics for bridges into 'usage', for use with
2505 * memory_report(). */
2507 bridge_get_memory_usage(struct simap *usage)
2514 ofproto_enumerate_types(&types);
2515 SSET_FOR_EACH (type, &types) {
2516 ofproto_type_get_memory_usage(type, usage);
2518 sset_destroy(&types);
2520 HMAP_FOR_EACH (br, node, &all_bridges) {
2521 ofproto_get_memory_usage(br->ofproto, usage);
2525 /* QoS unixctl user interface functions. */
2527 struct qos_unixctl_show_cbdata {
2529 struct iface *iface;
2533 qos_unixctl_show_queue(unsigned int queue_id,
2534 const struct smap *details,
2535 struct iface *iface,
2538 struct netdev_queue_stats stats;
2539 struct smap_node *node;
2542 ds_put_cstr(ds, "\n");
2544 ds_put_format(ds, "Queue %u:\n", queue_id);
2546 ds_put_cstr(ds, "Default:\n");
2549 SMAP_FOR_EACH (node, details) {
2550 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2553 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2555 if (stats.tx_packets != UINT64_MAX) {
2556 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2559 if (stats.tx_bytes != UINT64_MAX) {
2560 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2563 if (stats.tx_errors != UINT64_MAX) {
2564 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2567 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2568 queue_id, ovs_strerror(error));
2573 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2574 const char *argv[], void *aux OVS_UNUSED)
2576 struct ds ds = DS_EMPTY_INITIALIZER;
2577 struct smap smap = SMAP_INITIALIZER(&smap);
2578 struct iface *iface;
2580 struct smap_node *node;
2582 iface = iface_find(argv[1]);
2584 unixctl_command_reply_error(conn, "no such interface");
2588 netdev_get_qos(iface->netdev, &type, &smap);
2590 if (*type != '\0') {
2591 struct netdev_queue_dump dump;
2592 struct smap details;
2593 unsigned int queue_id;
2595 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2597 SMAP_FOR_EACH (node, &smap) {
2598 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2601 smap_init(&details);
2602 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2603 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2605 smap_destroy(&details);
2607 unixctl_command_reply(conn, ds_cstr(&ds));
2609 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2610 unixctl_command_reply_error(conn, ds_cstr(&ds));
2613 smap_destroy(&smap);
2617 /* Bridge reconfiguration functions. */
2619 bridge_create(const struct ovsrec_bridge *br_cfg)
2623 ovs_assert(!bridge_lookup(br_cfg->name));
2624 br = xzalloc(sizeof *br);
2626 br->name = xstrdup(br_cfg->name);
2627 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2630 /* Derive the default Ethernet address from the bridge's UUID. This should
2631 * be unique and it will be stable between ovs-vswitchd runs. */
2632 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2633 eth_addr_mark_random(br->default_ea);
2635 hmap_init(&br->ports);
2636 hmap_init(&br->ifaces);
2637 hmap_init(&br->iface_by_name);
2638 hmap_init(&br->mirrors);
2640 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2644 bridge_destroy(struct bridge *br)
2647 struct mirror *mirror, *next_mirror;
2648 struct port *port, *next_port;
2650 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2653 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2654 mirror_destroy(mirror);
2657 hmap_remove(&all_bridges, &br->node);
2658 ofproto_destroy(br->ofproto);
2659 hmap_destroy(&br->ifaces);
2660 hmap_destroy(&br->ports);
2661 hmap_destroy(&br->iface_by_name);
2662 hmap_destroy(&br->mirrors);
2669 static struct bridge *
2670 bridge_lookup(const char *name)
2674 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2675 if (!strcmp(br->name, name)) {
2682 /* Handle requests for a listing of all flows known by the OpenFlow
2683 * stack, including those normally hidden. */
2685 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2686 const char *argv[], void *aux OVS_UNUSED)
2691 br = bridge_lookup(argv[1]);
2693 unixctl_command_reply_error(conn, "Unknown bridge");
2698 ofproto_get_all_flows(br->ofproto, &results);
2700 unixctl_command_reply(conn, ds_cstr(&results));
2701 ds_destroy(&results);
2704 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2705 * connections and reconnect. If BRIDGE is not specified, then all bridges
2706 * drop their controller connections and reconnect. */
2708 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2709 const char *argv[], void *aux OVS_UNUSED)
2713 br = bridge_lookup(argv[1]);
2715 unixctl_command_reply_error(conn, "Unknown bridge");
2718 ofproto_reconnect_controllers(br->ofproto);
2720 HMAP_FOR_EACH (br, node, &all_bridges) {
2721 ofproto_reconnect_controllers(br->ofproto);
2724 unixctl_command_reply(conn, NULL);
2728 bridge_get_controllers(const struct bridge *br,
2729 struct ovsrec_controller ***controllersp)
2731 struct ovsrec_controller **controllers;
2732 size_t n_controllers;
2734 controllers = br->cfg->controller;
2735 n_controllers = br->cfg->n_controller;
2737 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2743 *controllersp = controllers;
2745 return n_controllers;
2749 bridge_collect_wanted_ports(struct bridge *br,
2750 const unsigned long int *splinter_vlans,
2751 struct shash *wanted_ports)
2755 shash_init(wanted_ports);
2757 for (i = 0; i < br->cfg->n_ports; i++) {
2758 const char *name = br->cfg->ports[i]->name;
2759 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2760 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2764 if (bridge_get_controllers(br, NULL)
2765 && !shash_find(wanted_ports, br->name)) {
2766 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2767 br->name, br->name);
2769 ovsrec_interface_init(&br->synth_local_iface);
2770 ovsrec_port_init(&br->synth_local_port);
2772 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2773 br->synth_local_port.n_interfaces = 1;
2774 br->synth_local_port.name = br->name;
2776 br->synth_local_iface.name = br->name;
2777 br->synth_local_iface.type = "internal";
2779 br->synth_local_ifacep = &br->synth_local_iface;
2781 shash_add(wanted_ports, br->name, &br->synth_local_port);
2784 if (splinter_vlans) {
2785 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2789 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2790 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2791 * 'br' needs to complete its configuration. */
2793 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2795 struct shash_node *port_node;
2796 struct port *port, *next;
2798 /* Get rid of deleted ports.
2799 * Get rid of deleted interfaces on ports that still exist. */
2800 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2801 port->cfg = shash_find_data(wanted_ports, port->name);
2805 port_del_ifaces(port);
2809 /* Update iface->cfg and iface->type in interfaces that still exist. */
2810 SHASH_FOR_EACH (port_node, wanted_ports) {
2811 const struct ovsrec_port *port = port_node->data;
2814 for (i = 0; i < port->n_interfaces; i++) {
2815 const struct ovsrec_interface *cfg = port->interfaces[i];
2816 struct iface *iface = iface_lookup(br, cfg->name);
2817 const char *type = iface_get_type(cfg, br->cfg);
2822 } else if (!strcmp(type, "null")) {
2823 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2824 " may be removed in February 2013. Please email"
2825 " dev@openvswitch.org with concerns.",
2828 /* We will add new interfaces later. */
2834 /* Initializes 'oc' appropriately as a management service controller for
2837 * The caller must free oc->target when it is no longer needed. */
2839 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2840 struct ofproto_controller *oc)
2842 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2843 oc->max_backoff = 0;
2844 oc->probe_interval = 60;
2845 oc->band = OFPROTO_OUT_OF_BAND;
2847 oc->burst_limit = 0;
2848 oc->enable_async_msgs = true;
2851 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2853 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2854 struct ofproto_controller *oc)
2858 oc->target = c->target;
2859 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2860 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2861 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2862 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2863 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2864 oc->burst_limit = (c->controller_burst_limit
2865 ? *c->controller_burst_limit : 0);
2866 oc->enable_async_msgs = (!c->enable_async_messages
2867 || *c->enable_async_messages);
2868 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2869 if (dscp < 0 || dscp > 63) {
2870 dscp = DSCP_DEFAULT;
2875 /* Configures the IP stack for 'br''s local interface properly according to the
2876 * configuration in 'c'. */
2878 bridge_configure_local_iface_netdev(struct bridge *br,
2879 struct ovsrec_controller *c)
2881 struct netdev *netdev;
2882 struct in_addr mask, gateway;
2884 struct iface *local_iface;
2887 /* If there's no local interface or no IP address, give up. */
2888 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2889 if (!local_iface || !c->local_ip || !inet_aton(c->local_ip, &ip)) {
2893 /* Bring up the local interface. */
2894 netdev = local_iface->netdev;
2895 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2897 /* Configure the IP address and netmask. */
2898 if (!c->local_netmask
2899 || !inet_aton(c->local_netmask, &mask)
2901 mask.s_addr = guess_netmask(ip.s_addr);
2903 if (!netdev_set_in4(netdev, ip, mask)) {
2904 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2905 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2908 /* Configure the default gateway. */
2909 if (c->local_gateway
2910 && inet_aton(c->local_gateway, &gateway)
2911 && gateway.s_addr) {
2912 if (!netdev_add_router(netdev, gateway)) {
2913 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2914 br->name, IP_ARGS(gateway.s_addr));
2919 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2920 * in either string are treated as equal to any number of slashes in the other,
2921 * e.g. "x///y" is equal to "x/y".
2923 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2924 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2925 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2926 * 'b' against a prefix of 'a'.
2929 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2931 const char *b_start = b;
2933 if (b - b_start >= b_stoplen) {
2935 } else if (*a != *b) {
2937 } else if (*a == '/') {
2938 a += strspn(a, "/");
2939 b += strspn(b, "/");
2940 } else if (*a == '\0') {
2950 bridge_configure_remotes(struct bridge *br,
2951 const struct sockaddr_in *managers, size_t n_managers)
2953 bool disable_in_band;
2955 struct ovsrec_controller **controllers;
2956 size_t n_controllers;
2958 enum ofproto_fail_mode fail_mode;
2960 struct ofproto_controller *ocs;
2964 /* Check if we should disable in-band control on this bridge. */
2965 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2968 /* Set OpenFlow queue ID for in-band control. */
2969 ofproto_set_in_band_queue(br->ofproto,
2970 smap_get_int(&br->cfg->other_config,
2971 "in-band-queue", -1));
2973 if (disable_in_band) {
2974 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2976 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2979 n_controllers = bridge_get_controllers(br, &controllers);
2981 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
2984 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
2985 for (i = 0; i < n_controllers; i++) {
2986 struct ovsrec_controller *c = controllers[i];
2988 if (!strncmp(c->target, "punix:", 6)
2989 || !strncmp(c->target, "unix:", 5)) {
2990 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2993 if (!strncmp(c->target, "unix:", 5)) {
2994 /* Connect to a listening socket */
2995 whitelist = xasprintf("unix:%s/", ovs_rundir());
2996 if (strchr(c->target, '/') &&
2997 !equal_pathnames(c->target, whitelist,
2998 strlen(whitelist))) {
2999 /* Absolute path specified, but not in ovs_rundir */
3000 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3001 "controller \"%s\" due to possibility for "
3002 "remote exploit. Instead, specify socket "
3003 "in whitelisted \"%s\" or connect to "
3004 "\"unix:%s/%s.mgmt\" (which is always "
3005 "available without special configuration).",
3006 br->name, c->target, whitelist,
3007 ovs_rundir(), br->name);
3012 whitelist = xasprintf("punix:%s/%s.controller",
3013 ovs_rundir(), br->name);
3014 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3015 /* Prevent remote ovsdb-server users from accessing
3016 * arbitrary Unix domain sockets and overwriting arbitrary
3018 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3019 "controller \"%s\" due to possibility of "
3020 "overwriting local files. Instead, specify "
3021 "whitelisted \"%s\" or connect to "
3022 "\"unix:%s/%s.mgmt\" (which is always "
3023 "available without special configuration).",
3024 br->name, c->target, whitelist,
3025 ovs_rundir(), br->name);
3034 bridge_configure_local_iface_netdev(br, c);
3035 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3036 if (disable_in_band) {
3037 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3042 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3043 bridge_get_allowed_versions(br));
3044 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3047 /* Set the fail-mode. */
3048 fail_mode = !br->cfg->fail_mode
3049 || !strcmp(br->cfg->fail_mode, "standalone")
3050 ? OFPROTO_FAIL_STANDALONE
3051 : OFPROTO_FAIL_SECURE;
3052 ofproto_set_fail_mode(br->ofproto, fail_mode);
3054 /* Configure OpenFlow controller connection snooping. */
3055 if (!ofproto_has_snoops(br->ofproto)) {
3059 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3060 ovs_rundir(), br->name));
3061 ofproto_set_snoops(br->ofproto, &snoops);
3062 sset_destroy(&snoops);
3067 bridge_configure_tables(struct bridge *br)
3069 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3073 n_tables = ofproto_get_n_tables(br->ofproto);
3075 for (i = 0; i < n_tables; i++) {
3076 struct ofproto_table_settings s;
3079 s.max_flows = UINT_MAX;
3082 s.n_prefix_fields = 0;
3083 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3085 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3086 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3089 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3090 s.max_flows = *cfg->flow_limit;
3092 if (cfg->overflow_policy
3093 && !strcmp(cfg->overflow_policy, "evict")) {
3095 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3096 for (k = 0; k < cfg->n_groups; k++) {
3097 const char *string = cfg->groups[k];
3100 msg = mf_parse_subfield__(&s.groups[k], &string);
3102 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3103 "'groups' (%s)", br->name, i, msg);
3105 } else if (*string) {
3106 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3107 "element '%s' contains trailing garbage",
3108 br->name, i, cfg->groups[k]);
3114 /* Prefix lookup fields. */
3115 s.n_prefix_fields = 0;
3116 for (k = 0; k < cfg->n_prefixes; k++) {
3117 const char *name = cfg->prefixes[k];
3118 const struct mf_field *mf = mf_from_name(name);
3120 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3124 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3125 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3126 "%s", br->name, name);
3129 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3130 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3131 "field not used: %s", br->name, name);
3134 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3136 if (s.n_prefix_fields > 0) {
3138 struct ds ds = DS_EMPTY_INITIALIZER;
3139 for (k = 0; k < s.n_prefix_fields; k++) {
3141 ds_put_char(&ds, ',');
3143 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3145 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3146 br->name, i, ds_cstr(&ds));
3151 ofproto_configure_table(br->ofproto, i, &s);
3155 for (; j < br->cfg->n_flow_tables; j++) {
3156 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3157 "%"PRId64" not supported by this datapath", br->name,
3158 br->cfg->key_flow_tables[j]);
3163 bridge_configure_dp_desc(struct bridge *br)
3165 ofproto_set_dp_desc(br->ofproto,
3166 smap_get(&br->cfg->other_config, "dp-desc"));
3169 /* Port functions. */
3171 static struct port *
3172 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3176 port = xzalloc(sizeof *port);
3178 port->name = xstrdup(cfg->name);
3180 list_init(&port->ifaces);
3182 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3186 /* Deletes interfaces from 'port' that are no longer configured for it. */
3188 port_del_ifaces(struct port *port)
3190 struct iface *iface, *next;
3191 struct sset new_ifaces;
3194 /* Collect list of new interfaces. */
3195 sset_init(&new_ifaces);
3196 for (i = 0; i < port->cfg->n_interfaces; i++) {
3197 const char *name = port->cfg->interfaces[i]->name;
3198 const char *type = port->cfg->interfaces[i]->type;
3199 if (strcmp(type, "null")) {
3200 sset_add(&new_ifaces, name);
3204 /* Get rid of deleted interfaces. */
3205 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3206 if (!sset_contains(&new_ifaces, iface->name)) {
3207 iface_destroy(iface);
3211 sset_destroy(&new_ifaces);
3215 port_destroy(struct port *port)
3218 struct bridge *br = port->bridge;
3219 struct iface *iface, *next;
3222 ofproto_bundle_unregister(br->ofproto, port);
3225 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3226 iface_destroy__(iface);
3229 hmap_remove(&br->ports, &port->hmap_node);
3235 static struct port *
3236 port_lookup(const struct bridge *br, const char *name)
3240 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3242 if (!strcmp(port->name, name)) {
3250 enable_lacp(struct port *port, bool *activep)
3252 if (!port->cfg->lacp) {
3253 /* XXX when LACP implementation has been sufficiently tested, enable by
3254 * default and make active on bonded ports. */
3256 } else if (!strcmp(port->cfg->lacp, "off")) {
3258 } else if (!strcmp(port->cfg->lacp, "active")) {
3261 } else if (!strcmp(port->cfg->lacp, "passive")) {
3265 VLOG_WARN("port %s: unknown LACP mode %s",
3266 port->name, port->cfg->lacp);
3271 static struct lacp_settings *
3272 port_configure_lacp(struct port *port, struct lacp_settings *s)
3274 const char *lacp_time, *system_id;
3277 if (!enable_lacp(port, &s->active)) {
3281 s->name = port->name;
3283 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3285 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3286 ETH_ADDR_SCAN_ARGS(s->id))) {
3287 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3288 " address.", port->name, system_id);
3292 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3295 if (eth_addr_is_zero(s->id)) {
3296 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3300 /* Prefer bondable links if unspecified. */
3301 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3303 s->priority = (priority > 0 && priority <= UINT16_MAX
3305 : UINT16_MAX - !list_is_short(&port->ifaces));
3307 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3308 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3310 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3311 "lacp-fallback-ab", false);
3317 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3319 int priority, portid, key;
3321 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3322 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3324 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3326 if (portid <= 0 || portid > UINT16_MAX) {
3327 portid = ofp_to_u16(iface->ofp_port);
3330 if (priority <= 0 || priority > UINT16_MAX) {
3331 priority = UINT16_MAX;
3334 if (key < 0 || key > UINT16_MAX) {
3338 s->name = iface->name;
3340 s->priority = priority;
3345 port_configure_bond(struct port *port, struct bond_settings *s)
3347 const char *detect_s;
3348 struct iface *iface;
3349 int miimon_interval;
3351 s->name = port->name;
3353 if (port->cfg->bond_mode) {
3354 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3355 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3356 port->name, port->cfg->bond_mode,
3357 bond_mode_to_string(s->balance));
3360 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3362 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3363 * active-backup. At some point we should remove this warning. */
3364 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3365 " in previous versions, the default bond_mode was"
3366 " balance-slb", port->name,
3367 bond_mode_to_string(s->balance));
3369 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3370 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3371 "please use another bond type or disable flood_vlans",
3375 miimon_interval = smap_get_int(&port->cfg->other_config,
3376 "bond-miimon-interval", 0);
3377 if (miimon_interval <= 0) {
3378 miimon_interval = 200;
3381 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3382 if (!detect_s || !strcmp(detect_s, "carrier")) {
3383 miimon_interval = 0;
3384 } else if (strcmp(detect_s, "miimon")) {
3385 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3386 "defaulting to carrier", port->name, detect_s);
3387 miimon_interval = 0;
3390 s->up_delay = MAX(0, port->cfg->bond_updelay);
3391 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3392 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3393 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3394 "bond-rebalance-interval", 10000);
3395 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3396 s->rebalance_interval = 1000;
3399 s->fake_iface = port->cfg->bond_fake_iface;
3401 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3402 "lacp-fallback-ab", false);
3404 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3405 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3409 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3410 * instead of obtaining it from the database. */
3412 port_is_synthetic(const struct port *port)
3414 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3417 /* Interface functions. */
3420 iface_is_internal(const struct ovsrec_interface *iface,
3421 const struct ovsrec_bridge *br)
3423 /* The local port and "internal" ports are always "internal". */
3424 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3427 /* Returns the correct network device type for interface 'iface' in bridge
3430 iface_get_type(const struct ovsrec_interface *iface,
3431 const struct ovsrec_bridge *br)
3435 /* The local port always has type "internal". Other ports take
3436 * their type from the database and default to "system" if none is
3438 if (iface_is_internal(iface, br)) {
3441 type = iface->type[0] ? iface->type : "system";
3444 return ofproto_port_open_type(br->datapath_type, type);
3448 iface_destroy__(struct iface *iface)
3451 struct port *port = iface->port;
3452 struct bridge *br = port->bridge;
3454 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3455 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3458 if (iface->ofp_port != OFPP_NONE) {
3459 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3462 list_remove(&iface->port_elem);
3463 hmap_remove(&br->iface_by_name, &iface->name_node);
3465 netdev_close(iface->netdev);
3473 iface_destroy(struct iface *iface)
3476 struct port *port = iface->port;
3478 iface_destroy__(iface);
3479 if (list_is_empty(&port->ifaces)) {
3485 static struct iface *
3486 iface_lookup(const struct bridge *br, const char *name)
3488 struct iface *iface;
3490 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3491 &br->iface_by_name) {
3492 if (!strcmp(iface->name, name)) {
3500 static struct iface *
3501 iface_find(const char *name)
3503 const struct bridge *br;
3505 HMAP_FOR_EACH (br, node, &all_bridges) {
3506 struct iface *iface = iface_lookup(br, name);
3515 static struct iface *
3516 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3518 struct iface *iface;
3520 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3522 if (iface->ofp_port == ofp_port) {
3529 /* Set Ethernet address of 'iface', if one is specified in the configuration
3532 iface_set_mac(struct iface *iface)
3534 uint8_t ea[ETH_ADDR_LEN];
3536 if (!strcmp(iface->type, "internal")
3537 && iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3538 if (iface->ofp_port == OFPP_LOCAL) {
3539 VLOG_ERR("interface %s: ignoring mac in Interface record "
3540 "(use Bridge record to set local port's mac)",
3542 } else if (eth_addr_is_multicast(ea)) {
3543 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3546 int error = netdev_set_etheraddr(iface->netdev, ea);
3548 VLOG_ERR("interface %s: setting MAC failed (%s)",
3549 iface->name, ovs_strerror(error));
3555 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3557 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3559 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3560 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3561 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3565 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3566 * sets the "ofport" field to -1.
3568 * This is appropriate when 'if_cfg''s interface cannot be created or is
3569 * otherwise invalid. */
3571 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3573 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3574 iface_set_ofport(if_cfg, OFPP_NONE);
3575 ovsrec_interface_set_status(if_cfg, NULL);
3576 ovsrec_interface_set_admin_state(if_cfg, NULL);
3577 ovsrec_interface_set_duplex(if_cfg, NULL);
3578 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3579 ovsrec_interface_set_link_state(if_cfg, NULL);
3580 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3581 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3582 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3583 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3584 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3585 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3586 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3587 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3592 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3594 union ovsdb_atom atom;
3596 atom.integer = target;
3597 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3601 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3603 struct ofpbuf queues_buf;
3605 ofpbuf_init(&queues_buf, 0);
3607 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3608 netdev_set_qos(iface->netdev, NULL, NULL);
3610 const struct ovsdb_datum *queues;
3611 struct netdev_queue_dump dump;
3612 unsigned int queue_id;
3613 struct smap details;
3617 /* Configure top-level Qos for 'iface'. */
3618 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3620 /* Deconfigure queues that were deleted. */
3621 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3623 smap_init(&details);
3624 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3625 if (!queue_ids_include(queues, queue_id)) {
3626 netdev_delete_queue(iface->netdev, queue_id);
3629 smap_destroy(&details);
3631 /* Configure queues for 'iface'. */
3633 for (i = 0; i < qos->n_queues; i++) {
3634 const struct ovsrec_queue *queue = qos->value_queues[i];
3635 unsigned int queue_id = qos->key_queues[i];
3637 if (queue_id == 0) {
3641 if (queue->n_dscp == 1) {
3642 struct ofproto_port_queue *port_queue;
3644 port_queue = ofpbuf_put_uninit(&queues_buf,
3645 sizeof *port_queue);
3646 port_queue->queue = queue_id;
3647 port_queue->dscp = queue->dscp[0];
3650 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3653 struct smap details;
3655 smap_init(&details);
3656 netdev_set_queue(iface->netdev, 0, &details);
3657 smap_destroy(&details);
3661 if (iface->ofp_port != OFPP_NONE) {
3662 const struct ofproto_port_queue *port_queues = queues_buf.data;
3663 size_t n_queues = queues_buf.size / sizeof *port_queues;
3665 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3666 port_queues, n_queues);
3669 netdev_set_policing(iface->netdev,
3670 iface->cfg->ingress_policing_rate,
3671 iface->cfg->ingress_policing_burst);
3673 ofpbuf_uninit(&queues_buf);
3677 iface_configure_cfm(struct iface *iface)
3679 const struct ovsrec_interface *cfg = iface->cfg;
3680 const char *opstate_str;
3681 const char *cfm_ccm_vlan;
3682 struct cfm_settings s;
3683 struct smap netdev_args;
3685 if (!cfg->n_cfm_mpid) {
3686 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3690 s.check_tnl_key = false;
3691 smap_init(&netdev_args);
3692 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3693 const char *key = smap_get(&netdev_args, "key");
3694 const char *in_key = smap_get(&netdev_args, "in_key");
3696 s.check_tnl_key = (key && !strcmp(key, "flow"))
3697 || (in_key && !strcmp(in_key, "flow"));
3699 smap_destroy(&netdev_args);
3701 s.mpid = *cfg->cfm_mpid;
3702 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3703 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3704 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3706 if (s.interval <= 0) {
3710 if (!cfm_ccm_vlan) {
3712 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3713 s.ccm_vlan = CFM_RANDOM_VLAN;
3715 s.ccm_vlan = atoi(cfm_ccm_vlan);
3716 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3721 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3723 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3725 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3726 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3728 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3731 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3732 * instead of obtaining it from the database. */
3734 iface_is_synthetic(const struct iface *iface)
3736 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3740 iface_validate_ofport__(size_t n, int64_t *ofport)
3742 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3743 ? u16_to_ofp(*ofport)
3748 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3750 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3754 iface_pick_ofport(const struct ovsrec_interface *cfg)
3756 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3757 return (requested_ofport != OFPP_NONE
3759 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3762 /* Port mirroring. */
3764 static struct mirror *
3765 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3769 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3770 if (uuid_equals(uuid, &m->uuid)) {
3778 bridge_configure_mirrors(struct bridge *br)
3780 const struct ovsdb_datum *mc;
3781 unsigned long *flood_vlans;
3782 struct mirror *m, *next;
3785 /* Get rid of deleted mirrors. */
3786 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3787 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3788 union ovsdb_atom atom;
3790 atom.uuid = m->uuid;
3791 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3796 /* Add new mirrors and reconfigure existing ones. */
3797 for (i = 0; i < br->cfg->n_mirrors; i++) {
3798 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3799 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3801 m = mirror_create(br, cfg);
3804 if (!mirror_configure(m)) {
3809 /* Update flooded vlans (for RSPAN). */
3810 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3811 br->cfg->n_flood_vlans);
3812 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3813 bitmap_free(flood_vlans);
3816 static struct mirror *
3817 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3821 m = xzalloc(sizeof *m);
3822 m->uuid = cfg->header_.uuid;
3823 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3825 m->name = xstrdup(cfg->name);
3831 mirror_destroy(struct mirror *m)
3834 struct bridge *br = m->bridge;
3837 ofproto_mirror_unregister(br->ofproto, m);
3840 hmap_remove(&br->mirrors, &m->hmap_node);
3847 mirror_collect_ports(struct mirror *m,
3848 struct ovsrec_port **in_ports, int n_in_ports,
3849 void ***out_portsp, size_t *n_out_portsp)
3851 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3852 size_t n_out_ports = 0;
3855 for (i = 0; i < n_in_ports; i++) {
3856 const char *name = in_ports[i]->name;
3857 struct port *port = port_lookup(m->bridge, name);
3859 out_ports[n_out_ports++] = port;
3861 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3862 "port %s", m->bridge->name, m->name, name);
3865 *out_portsp = out_ports;
3866 *n_out_portsp = n_out_ports;
3870 mirror_configure(struct mirror *m)
3872 const struct ovsrec_mirror *cfg = m->cfg;
3873 struct ofproto_mirror_settings s;
3876 if (strcmp(cfg->name, m->name)) {
3878 m->name = xstrdup(cfg->name);
3882 /* Get output port or VLAN. */
3883 if (cfg->output_port) {
3884 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3885 if (!s.out_bundle) {
3886 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3887 m->bridge->name, m->name);
3890 s.out_vlan = UINT16_MAX;
3892 if (cfg->output_vlan) {
3893 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3894 "output vlan; ignoring output vlan",
3895 m->bridge->name, m->name);
3897 } else if (cfg->output_vlan) {
3898 /* The database should prevent invalid VLAN values. */
3899 s.out_bundle = NULL;
3900 s.out_vlan = *cfg->output_vlan;
3902 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3903 m->bridge->name, m->name);
3907 /* Get port selection. */
3908 if (cfg->select_all) {
3909 size_t n_ports = hmap_count(&m->bridge->ports);
3910 void **ports = xmalloc(n_ports * sizeof *ports);
3915 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3925 /* Get ports, dropping ports that don't exist.
3926 * The IDL ensures that there are no duplicates. */
3927 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3928 &s.srcs, &s.n_srcs);
3929 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3930 &s.dsts, &s.n_dsts);
3933 /* Get VLAN selection. */
3934 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3937 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3940 if (s.srcs != s.dsts) {
3949 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3951 * This is deprecated. It is only for compatibility with broken device drivers
3952 * in old versions of Linux that do not properly support VLANs when VLAN
3953 * devices are not used. When broken device drivers are no longer in
3954 * widespread use, we will delete these interfaces. */
3956 static struct ovsrec_port **recs;
3957 static size_t n_recs, allocated_recs;
3959 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3960 * splinters are reconfigured. */
3962 register_rec(struct ovsrec_port *rec)
3964 if (n_recs >= allocated_recs) {
3965 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3967 recs[n_recs++] = rec;
3970 /* Frees all of the ports registered with register_reg(). */
3972 free_registered_recs(void)
3976 for (i = 0; i < n_recs; i++) {
3977 struct ovsrec_port *port = recs[i];
3980 for (j = 0; j < port->n_interfaces; j++) {
3981 struct ovsrec_interface *iface = port->interfaces[j];
3986 smap_destroy(&port->other_config);
3987 free(port->interfaces);
3995 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
3998 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4000 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4004 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4007 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4008 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4009 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4012 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4013 * use, returns NULL.
4015 * Updates 'vlan_splinters_enabled_anywhere'. */
4016 static unsigned long int *
4017 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4019 unsigned long int *splinter_vlans;
4020 struct sset splinter_ifaces;
4021 const char *real_dev_name;
4022 struct shash *real_devs;
4023 struct shash_node *node;
4027 /* Free space allocated for synthesized ports and interfaces, since we're
4028 * in the process of reconstructing all of them. */
4029 free_registered_recs();
4031 splinter_vlans = bitmap_allocate(4096);
4032 sset_init(&splinter_ifaces);
4033 vlan_splinters_enabled_anywhere = false;
4034 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4035 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4038 for (j = 0; j < br_cfg->n_ports; j++) {
4039 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4042 for (k = 0; k < port_cfg->n_interfaces; k++) {
4043 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4045 if (vlan_splinters_is_enabled(iface_cfg)) {
4046 vlan_splinters_enabled_anywhere = true;
4047 sset_add(&splinter_ifaces, iface_cfg->name);
4048 vlan_bitmap_from_array__(port_cfg->trunks,
4054 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4055 bitmap_set1(splinter_vlans, *port_cfg->tag);
4060 if (!vlan_splinters_enabled_anywhere) {
4061 free(splinter_vlans);
4062 sset_destroy(&splinter_ifaces);
4066 HMAP_FOR_EACH (br, node, &all_bridges) {
4068 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4072 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4073 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4074 * device to be created for it. */
4075 bitmap_set0(splinter_vlans, 0);
4076 bitmap_set0(splinter_vlans, 4095);
4078 /* Delete all VLAN devices that we don't need. */
4080 real_devs = vlandev_get_real_devs();
4081 SHASH_FOR_EACH (node, real_devs) {
4082 const struct vlan_real_dev *real_dev = node->data;
4083 const struct vlan_dev *vlan_dev;
4084 bool real_dev_has_splinters;
4086 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4088 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4089 if (!real_dev_has_splinters
4090 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4091 struct netdev *netdev;
4093 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4094 if (!netdev_get_in4(netdev, NULL, NULL) ||
4095 !netdev_get_in6(netdev, NULL)) {
4096 /* It has an IP address configured, so we don't own
4097 * it. Don't delete it. */
4099 vlandev_del(vlan_dev->name);
4101 netdev_close(netdev);
4108 /* Add all VLAN devices that we need. */
4109 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4112 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4113 if (!vlandev_get_name(real_dev_name, vid)) {
4114 vlandev_add(real_dev_name, vid);
4121 sset_destroy(&splinter_ifaces);
4123 if (bitmap_scan(splinter_vlans, 0, 4096) >= 4096) {
4124 free(splinter_vlans);
4127 return splinter_vlans;
4130 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4133 configure_splinter_port(struct port *port)
4135 struct ofproto *ofproto = port->bridge->ofproto;
4136 ofp_port_t realdev_ofp_port;
4137 const char *realdev_name;
4138 struct iface *vlandev, *realdev;
4140 ofproto_bundle_unregister(port->bridge->ofproto, port);
4142 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4145 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4146 realdev = iface_lookup(port->bridge, realdev_name);
4147 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4149 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4153 static struct ovsrec_port *
4154 synthesize_splinter_port(const char *real_dev_name,
4155 const char *vlan_dev_name, int vid)
4157 struct ovsrec_interface *iface;
4158 struct ovsrec_port *port;
4160 iface = xmalloc(sizeof *iface);
4161 ovsrec_interface_init(iface);
4162 iface->name = xstrdup(vlan_dev_name);
4163 iface->type = "system";
4165 port = xmalloc(sizeof *port);
4166 ovsrec_port_init(port);
4167 port->interfaces = xmemdup(&iface, sizeof iface);
4168 port->n_interfaces = 1;
4169 port->name = xstrdup(vlan_dev_name);
4170 port->vlan_mode = "splinter";
4171 port->tag = xmalloc(sizeof *port->tag);
4174 smap_add(&port->other_config, "realdev", real_dev_name);
4180 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4181 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4182 * 1-bit in the 'splinter_vlans' bitmap. */
4184 add_vlan_splinter_ports(struct bridge *br,
4185 const unsigned long int *splinter_vlans,
4186 struct shash *ports)
4190 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4191 * we're modifying 'ports'. */
4192 for (i = 0; i < br->cfg->n_ports; i++) {
4193 const char *name = br->cfg->ports[i]->name;
4194 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4197 for (j = 0; j < port_cfg->n_interfaces; j++) {
4198 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4200 if (vlan_splinters_is_enabled(iface_cfg)) {
4201 const char *real_dev_name;
4204 real_dev_name = iface_cfg->name;
4205 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4206 const char *vlan_dev_name;
4208 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4210 && !shash_find(ports, vlan_dev_name)) {
4211 shash_add(ports, vlan_dev_name,
4212 synthesize_splinter_port(
4213 real_dev_name, vlan_dev_name, vid));
4222 mirror_refresh_stats(struct mirror *m)
4224 struct ofproto *ofproto = m->bridge->ofproto;
4225 uint64_t tx_packets, tx_bytes;
4228 size_t stat_cnt = 0;
4230 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4231 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4235 if (tx_packets != UINT64_MAX) {
4236 keys[stat_cnt] = "tx_packets";
4237 values[stat_cnt] = tx_packets;
4240 if (tx_bytes != UINT64_MAX) {
4241 keys[stat_cnt] = "tx_bytes";
4242 values[stat_cnt] = tx_bytes;
4246 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);