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 /* Set to true to allow experimental use of OpenFlow 1.4.
169 * This is false initially because OpenFlow 1.4 is not yet safe to use: it can
170 * abort due to unimplemented features. */
171 static bool allow_of14;
173 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
174 * expensive. This can cause bridge_reconfigure() to take a long time during
175 * which no other work can be done. To deal with this problem, we limit port
176 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
177 * call to bridge_reconfigure(). If there is more work to do after the limit
178 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
179 * This allows the rest of the code to catch up on important things like
180 * forwarding packets. */
181 #define OFP_PORT_ACTION_WINDOW 10
183 static void add_del_bridges(const struct ovsrec_open_vswitch *);
184 static void bridge_run__(void);
185 static void bridge_create(const struct ovsrec_bridge *);
186 static void bridge_destroy(struct bridge *);
187 static struct bridge *bridge_lookup(const char *name);
188 static unixctl_cb_func bridge_unixctl_dump_flows;
189 static unixctl_cb_func bridge_unixctl_reconnect;
190 static size_t bridge_get_controllers(const struct bridge *br,
191 struct ovsrec_controller ***controllersp);
192 static void bridge_collect_wanted_ports(struct bridge *,
193 const unsigned long *splinter_vlans,
194 struct shash *wanted_ports);
195 static void bridge_delete_ofprotos(void);
196 static void bridge_delete_or_reconfigure_ports(struct bridge *);
197 static void bridge_del_ports(struct bridge *,
198 const struct shash *wanted_ports);
199 static void bridge_add_ports(struct bridge *,
200 const struct shash *wanted_ports);
202 static void bridge_configure_datapath_id(struct bridge *);
203 static void bridge_configure_netflow(struct bridge *);
204 static void bridge_configure_forward_bpdu(struct bridge *);
205 static void bridge_configure_mac_table(struct bridge *);
206 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
207 static void bridge_configure_ipfix(struct bridge *);
208 static void bridge_configure_stp(struct bridge *);
209 static void bridge_configure_tables(struct bridge *);
210 static void bridge_configure_dp_desc(struct bridge *);
211 static void bridge_configure_remotes(struct bridge *,
212 const struct sockaddr_in *managers,
214 static void bridge_pick_local_hw_addr(struct bridge *,
215 uint8_t ea[ETH_ADDR_LEN],
216 struct iface **hw_addr_iface);
217 static uint64_t bridge_pick_datapath_id(struct bridge *,
218 const uint8_t bridge_ea[ETH_ADDR_LEN],
219 struct iface *hw_addr_iface);
220 static uint64_t dpid_from_hash(const void *, size_t nbytes);
221 static bool bridge_has_bond_fake_iface(const struct bridge *,
223 static bool port_is_bond_fake_iface(const struct port *);
225 static unixctl_cb_func qos_unixctl_show;
227 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
228 static void port_del_ifaces(struct port *);
229 static void port_destroy(struct port *);
230 static struct port *port_lookup(const struct bridge *, const char *name);
231 static void port_configure(struct port *);
232 static struct lacp_settings *port_configure_lacp(struct port *,
233 struct lacp_settings *);
234 static void port_configure_bond(struct port *, struct bond_settings *);
235 static bool port_is_synthetic(const struct port *);
237 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
238 static void run_system_stats(void);
240 static void bridge_configure_mirrors(struct bridge *);
241 static struct mirror *mirror_create(struct bridge *,
242 const struct ovsrec_mirror *);
243 static void mirror_destroy(struct mirror *);
244 static bool mirror_configure(struct mirror *);
245 static void mirror_refresh_stats(struct mirror *);
247 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
248 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
249 const struct ovsrec_port *);
250 static bool iface_is_internal(const struct ovsrec_interface *iface,
251 const struct ovsrec_bridge *br);
252 static const char *iface_get_type(const struct ovsrec_interface *,
253 const struct ovsrec_bridge *);
254 static void iface_destroy(struct iface *);
255 static void iface_destroy__(struct iface *);
256 static struct iface *iface_lookup(const struct bridge *, const char *name);
257 static struct iface *iface_find(const char *name);
258 static struct iface *iface_from_ofp_port(const struct bridge *,
259 ofp_port_t ofp_port);
260 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
261 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
262 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg);
263 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
264 static void iface_configure_cfm(struct iface *);
265 static void iface_refresh_cfm_stats(struct iface *);
266 static void iface_refresh_stats(struct iface *);
267 static void iface_refresh_status(struct iface *);
268 static bool iface_is_synthetic(const struct iface *);
269 static ofp_port_t iface_get_requested_ofp_port(
270 const struct ovsrec_interface *);
271 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
273 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
275 * This is deprecated. It is only for compatibility with broken device drivers
276 * in old versions of Linux that do not properly support VLANs when VLAN
277 * devices are not used. When broken device drivers are no longer in
278 * widespread use, we will delete these interfaces. */
280 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
281 static bool vlan_splinters_enabled_anywhere;
283 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
284 static unsigned long int *collect_splinter_vlans(
285 const struct ovsrec_open_vswitch *);
286 static void configure_splinter_port(struct port *);
287 static void add_vlan_splinter_ports(struct bridge *,
288 const unsigned long int *splinter_vlans,
289 struct shash *ports);
292 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
294 struct shash iface_hints;
295 static bool initialized = false;
302 shash_init(&iface_hints);
305 for (i = 0; i < cfg->n_bridges; i++) {
306 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
309 for (j = 0; j < br_cfg->n_ports; j++) {
310 struct ovsrec_port *port_cfg = br_cfg->ports[j];
313 for (k = 0; k < port_cfg->n_interfaces; k++) {
314 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
315 struct iface_hint *iface_hint;
317 iface_hint = xmalloc(sizeof *iface_hint);
318 iface_hint->br_name = br_cfg->name;
319 iface_hint->br_type = br_cfg->datapath_type;
320 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
322 shash_add(&iface_hints, if_cfg->name, iface_hint);
328 ofproto_init(&iface_hints);
330 shash_destroy_free_data(&iface_hints);
334 /* Public functions. */
336 /* Initializes the bridge module, configuring it to obtain its configuration
337 * from an OVSDB server accessed over 'remote', which should be a string in a
338 * form acceptable to ovsdb_idl_create(). */
340 bridge_init(const char *remote)
342 /* Create connection to database. */
343 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
344 idl_seqno = ovsdb_idl_get_seqno(idl);
345 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
346 ovsdb_idl_verify_write_only(idl);
348 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
349 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
350 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
351 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
352 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
353 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
354 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
356 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
357 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
358 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
360 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
361 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
362 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
364 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
365 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
366 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
367 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
368 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
369 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
370 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
371 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
372 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
373 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
374 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
375 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
376 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
377 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
378 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
379 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
380 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
381 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
382 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
383 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
385 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
386 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
387 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
388 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
390 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
392 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
394 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
395 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
397 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
398 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
399 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
400 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
402 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
403 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
404 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
405 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
406 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
408 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
410 /* Register unixctl commands. */
411 unixctl_command_register("qos/show", "interface", 1, 1,
412 qos_unixctl_show, NULL);
413 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
414 bridge_unixctl_dump_flows, NULL);
415 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
416 bridge_unixctl_reconnect, NULL);
426 struct bridge *br, *next_br;
428 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
431 ovsdb_idl_destroy(idl);
434 /* Enables use of OpenFlow 1.4. This is off by default because OpenFlow 1.4 is
435 * not yet safe to use: it can abort due to unimplemented features. */
437 bridge_enable_of14(void)
442 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
443 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
444 * responsible for freeing '*managersp' (with free()).
446 * You may be asking yourself "why does ovs-vswitchd care?", because
447 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
448 * should not be and in fact is not directly involved in that. But
449 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
450 * it has to tell in-band control where the managers are to enable that.
451 * (Thus, only managers connected in-band are collected.)
454 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
455 struct sockaddr_in **managersp, size_t *n_managersp)
457 struct sockaddr_in *managers = NULL;
458 size_t n_managers = 0;
462 /* Collect all of the potential targets from the "targets" columns of the
463 * rows pointed to by "manager_options", excluding any that are
466 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
467 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
469 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
470 sset_find_and_delete(&targets, m->target);
472 sset_add(&targets, m->target);
476 /* Now extract the targets' IP addresses. */
477 if (!sset_is_empty(&targets)) {
480 managers = xmalloc(sset_count(&targets) * sizeof *managers);
481 SSET_FOR_EACH (target, &targets) {
482 struct sockaddr_storage ss;
484 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
486 && ss.ss_family == AF_INET) {
487 managers[n_managers++] = *(struct sockaddr_in *) &ss;
491 sset_destroy(&targets);
493 *managersp = managers;
494 *n_managersp = n_managers;
498 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
500 unsigned long int *splinter_vlans;
501 struct sockaddr_in *managers;
502 struct bridge *br, *next;
503 int sflow_bridge_number;
506 COVERAGE_INC(bridge_reconfigure);
508 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
509 OFPROTO_FLOW_LIMIT_DEFAULT));
510 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
511 OFPROTO_MAX_IDLE_DEFAULT));
514 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
515 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
517 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
518 * to 'ovs_cfg', with only very minimal configuration otherwise.
520 * This is mostly an update to bridge data structures. Nothing is pushed
521 * down to ofproto or lower layers. */
522 add_del_bridges(ovs_cfg);
523 splinter_vlans = collect_splinter_vlans(ovs_cfg);
524 HMAP_FOR_EACH (br, node, &all_bridges) {
525 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
526 bridge_del_ports(br, &br->wanted_ports);
528 free(splinter_vlans);
530 /* Start pushing configuration changes down to the ofproto layer:
532 * - Delete ofprotos that are no longer configured.
534 * - Delete ports that are no longer configured.
536 * - Reconfigure existing ports to their desired configurations, or
537 * delete them if not possible.
539 * We have to do all the deletions before we can do any additions, because
540 * the ports to be added might require resources that will be freed up by
541 * deletions (they might especially overlap in name). */
542 bridge_delete_ofprotos();
543 HMAP_FOR_EACH (br, node, &all_bridges) {
545 bridge_delete_or_reconfigure_ports(br);
549 /* Finish pushing configuration changes to the ofproto layer:
551 * - Create ofprotos that are missing.
553 * - Add ports that are missing. */
554 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
558 error = ofproto_create(br->name, br->type, &br->ofproto);
560 VLOG_ERR("failed to create bridge %s: %s", br->name,
561 ovs_strerror(error));
562 shash_destroy(&br->wanted_ports);
567 HMAP_FOR_EACH (br, node, &all_bridges) {
568 bridge_add_ports(br, &br->wanted_ports);
569 shash_destroy(&br->wanted_ports);
572 reconfigure_system_stats(ovs_cfg);
574 /* Complete the configuration. */
575 sflow_bridge_number = 0;
576 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
577 HMAP_FOR_EACH (br, node, &all_bridges) {
580 /* We need the datapath ID early to allow LACP ports to use it as the
581 * default system ID. */
582 bridge_configure_datapath_id(br);
584 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
587 port_configure(port);
589 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
590 iface_set_ofport(iface->cfg, iface->ofp_port);
591 iface_configure_cfm(iface);
592 iface_configure_qos(iface, port->cfg->qos);
593 iface_set_mac(br, port, iface);
594 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
598 bridge_configure_mirrors(br);
599 bridge_configure_forward_bpdu(br);
600 bridge_configure_mac_table(br);
601 bridge_configure_remotes(br, managers, n_managers);
602 bridge_configure_netflow(br);
603 bridge_configure_sflow(br, &sflow_bridge_number);
604 bridge_configure_ipfix(br);
605 bridge_configure_stp(br);
606 bridge_configure_tables(br);
607 bridge_configure_dp_desc(br);
609 if (smap_get(&br->cfg->other_config, "flow-eviction-threshold")) {
610 /* XXX: Remove this warning message eventually. */
611 VLOG_WARN_ONCE("As of June 2013, flow-eviction-threshold has been"
612 " moved to the Open_vSwitch table. Ignoring its"
613 " setting in the bridge table.");
618 /* The ofproto-dpif provider does some final reconfiguration in its
619 * ->type_run() function. We have to call it before notifying the database
620 * client that reconfiguration is complete, otherwise there is a very
621 * narrow race window in which e.g. ofproto/trace will not recognize the
622 * new configuration (sometimes this causes unit test failures). */
626 /* Delete ofprotos which aren't configured or have the wrong type. Create
627 * ofprotos which don't exist but need to. */
629 bridge_delete_ofprotos(void)
636 /* Delete ofprotos with no bridge or with the wrong type. */
639 ofproto_enumerate_types(&types);
640 SSET_FOR_EACH (type, &types) {
643 ofproto_enumerate_names(type, &names);
644 SSET_FOR_EACH (name, &names) {
645 br = bridge_lookup(name);
646 if (!br || strcmp(type, br->type)) {
647 ofproto_delete(name, type);
651 sset_destroy(&names);
652 sset_destroy(&types);
656 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
658 if (*n >= *allocated) {
659 ports = x2nrealloc(ports, allocated, sizeof *ports);
661 ports[(*n)++] = port;
666 bridge_delete_or_reconfigure_ports(struct bridge *br)
668 struct ofproto_port ofproto_port;
669 struct ofproto_port_dump dump;
671 struct sset ofproto_ports;
672 struct port *port, *port_next;
674 /* List of "ofp_port"s to delete. We make a list instead of deleting them
675 * right away because ofproto implementations aren't necessarily able to
676 * iterate through a changing list of ports in an entirely robust way. */
683 sset_init(&ofproto_ports);
685 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
686 * that are not configured in the database. (This commonly happens when
687 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
689 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
690 * that have the wrong OpenFlow port number (and arrange to add them back
691 * with the correct OpenFlow port number). */
692 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
693 ofp_port_t requested_ofp_port;
696 sset_add(&ofproto_ports, ofproto_port.name);
698 iface = iface_lookup(br, ofproto_port.name);
700 /* No such iface is configured, so we should delete this
703 * As a corner case exception, keep the port if it's a bond fake
705 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
706 && !strcmp(ofproto_port.type, "internal")) {
712 if (strcmp(ofproto_port.type, iface->type)
713 || netdev_set_config(iface->netdev, &iface->cfg->options)) {
714 /* The interface is the wrong type or can't be configured.
719 /* If the requested OpenFlow port for 'iface' changed, and it's not
720 * already the correct port, then we might want to temporarily delete
721 * this interface, so we can add it back again with the new OpenFlow
723 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
724 if (iface->ofp_port != OFPP_LOCAL &&
725 requested_ofp_port != OFPP_NONE &&
726 requested_ofp_port != iface->ofp_port) {
727 ofp_port_t victim_request;
728 struct iface *victim;
730 /* Check for an existing OpenFlow port currently occupying
731 * 'iface''s requested port number. If there isn't one, then
732 * delete this port. Otherwise we need to consider further. */
733 victim = iface_from_ofp_port(br, requested_ofp_port);
738 /* 'victim' is a port currently using 'iface''s requested port
739 * number. Unless 'victim' specifically requested that port
740 * number, too, then we can delete both 'iface' and 'victim'
741 * temporarily. (We'll add both of them back again later with new
742 * OpenFlow port numbers.)
744 * If 'victim' did request port number 'requested_ofp_port', just
745 * like 'iface', then that's a configuration inconsistency that we
746 * can't resolve. We might as well let it keep its current port
748 victim_request = iface_get_requested_ofp_port(victim->cfg);
749 if (victim_request != requested_ofp_port) {
750 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
751 iface_destroy(victim);
760 iface_destroy(iface);
761 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
763 for (i = 0; i < n; i++) {
764 ofproto_port_del(br->ofproto, del[i]);
768 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
769 * that we didn't see when we iterated through the datapath, i.e. ports
770 * that disappeared underneath use. This is an unusual situation, but it
771 * can happen in some cases:
773 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
774 * idea but could happen).
776 * - The port represented a device that disappeared, e.g. a tuntap
777 * device destroyed via "tunctl -d", a physical Ethernet device
778 * whose module was just unloaded via "rmmod", or a virtual NIC for a
779 * VM whose VM was just terminated. */
780 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
781 struct iface *iface, *iface_next;
783 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
784 if (!sset_contains(&ofproto_ports, iface->name)) {
785 iface_destroy__(iface);
789 if (list_is_empty(&port->ifaces)) {
793 sset_destroy(&ofproto_ports);
797 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
798 bool with_requested_port)
800 struct shash_node *port_node;
802 SHASH_FOR_EACH (port_node, wanted_ports) {
803 const struct ovsrec_port *port_cfg = port_node->data;
806 for (i = 0; i < port_cfg->n_interfaces; i++) {
807 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
808 ofp_port_t requested_ofp_port;
810 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
811 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
812 struct iface *iface = iface_lookup(br, iface_cfg->name);
815 iface_create(br, iface_cfg, port_cfg);
823 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
825 /* First add interfaces that request a particular port number. */
826 bridge_add_ports__(br, wanted_ports, true);
828 /* Then add interfaces that want automatic port number assignment.
829 * We add these afterward to avoid accidentally taking a specifically
830 * requested port number. */
831 bridge_add_ports__(br, wanted_ports, false);
835 port_configure(struct port *port)
837 const struct ovsrec_port *cfg = port->cfg;
838 struct bond_settings bond_settings;
839 struct lacp_settings lacp_settings;
840 struct ofproto_bundle_settings s;
843 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
844 configure_splinter_port(port);
853 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
854 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
855 s.slaves[s.n_slaves++] = iface->ofp_port;
860 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
864 /* Get VLAN trunks. */
867 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
871 if (cfg->vlan_mode) {
872 if (!strcmp(cfg->vlan_mode, "access")) {
873 s.vlan_mode = PORT_VLAN_ACCESS;
874 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
875 s.vlan_mode = PORT_VLAN_TRUNK;
876 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
877 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
878 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
879 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
881 /* This "can't happen" because ovsdb-server should prevent it. */
882 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
883 "back to trunk mode", port->name, cfg->vlan_mode);
884 s.vlan_mode = PORT_VLAN_TRUNK;
888 s.vlan_mode = PORT_VLAN_ACCESS;
890 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
894 s.vlan_mode = PORT_VLAN_TRUNK;
897 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
900 /* Get LACP settings. */
901 s.lacp = port_configure_lacp(port, &lacp_settings);
905 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
906 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
907 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
910 s.lacp_slaves = NULL;
913 /* Get bond settings. */
914 if (s.n_slaves > 1) {
915 s.bond = &bond_settings;
916 port_configure_bond(port, &bond_settings);
919 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
920 netdev_set_miimon_interval(iface->netdev, 0);
925 ofproto_bundle_register(port->bridge->ofproto, port, &s);
933 /* Pick local port hardware address and datapath ID for 'br'. */
935 bridge_configure_datapath_id(struct bridge *br)
937 uint8_t ea[ETH_ADDR_LEN];
939 struct iface *local_iface;
940 struct iface *hw_addr_iface;
943 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
944 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
946 int error = netdev_set_etheraddr(local_iface->netdev, ea);
948 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
949 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
950 "Ethernet address: %s",
951 br->name, ovs_strerror(error));
954 memcpy(br->ea, ea, ETH_ADDR_LEN);
956 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
957 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
958 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
959 ofproto_set_datapath_id(br->ofproto, dpid);
962 dpid_string = xasprintf("%016"PRIx64, dpid);
963 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
967 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
970 bridge_get_allowed_versions(struct bridge *br)
972 uint32_t allowed_versions;
974 if (!br->cfg->n_protocols)
977 allowed_versions = ofputil_versions_from_strings(br->cfg->protocols,
978 br->cfg->n_protocols);
980 allowed_versions &= ~(1u << OFP14_VERSION);
982 return allowed_versions;
985 /* Set NetFlow configuration on 'br'. */
987 bridge_configure_netflow(struct bridge *br)
989 struct ovsrec_netflow *cfg = br->cfg->netflow;
990 struct netflow_options opts;
993 ofproto_set_netflow(br->ofproto, NULL);
997 memset(&opts, 0, sizeof opts);
999 /* Get default NetFlow configuration from datapath.
1000 * Apply overrides from 'cfg'. */
1001 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1002 if (cfg->engine_type) {
1003 opts.engine_type = *cfg->engine_type;
1005 if (cfg->engine_id) {
1006 opts.engine_id = *cfg->engine_id;
1009 /* Configure active timeout interval. */
1010 opts.active_timeout = cfg->active_timeout;
1011 if (!opts.active_timeout) {
1012 opts.active_timeout = -1;
1013 } else if (opts.active_timeout < 0) {
1014 VLOG_WARN("bridge %s: active timeout interval set to negative "
1015 "value, using default instead (%d seconds)", br->name,
1016 NF_ACTIVE_TIMEOUT_DEFAULT);
1017 opts.active_timeout = -1;
1020 /* Add engine ID to interface number to disambiguate bridgs? */
1021 opts.add_id_to_iface = cfg->add_id_to_interface;
1022 if (opts.add_id_to_iface) {
1023 if (opts.engine_id > 0x7f) {
1024 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1025 "another vswitch, choose an engine id less than 128",
1028 if (hmap_count(&br->ports) > 508) {
1029 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1030 "another port when more than 508 ports are used",
1036 sset_init(&opts.collectors);
1037 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1040 if (ofproto_set_netflow(br->ofproto, &opts)) {
1041 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1043 sset_destroy(&opts.collectors);
1046 /* Set sFlow configuration on 'br'. */
1048 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1050 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1051 struct ovsrec_controller **controllers;
1052 struct ofproto_sflow_options oso;
1053 size_t n_controllers;
1057 ofproto_set_sflow(br->ofproto, NULL);
1061 memset(&oso, 0, sizeof oso);
1063 sset_init(&oso.targets);
1064 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1066 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1067 if (cfg->sampling) {
1068 oso.sampling_rate = *cfg->sampling;
1071 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1073 oso.polling_interval = *cfg->polling;
1076 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1078 oso.header_len = *cfg->header;
1081 oso.sub_id = (*sflow_bridge_number)++;
1082 oso.agent_device = cfg->agent;
1084 oso.control_ip = NULL;
1085 n_controllers = bridge_get_controllers(br, &controllers);
1086 for (i = 0; i < n_controllers; i++) {
1087 if (controllers[i]->local_ip) {
1088 oso.control_ip = controllers[i]->local_ip;
1092 ofproto_set_sflow(br->ofproto, &oso);
1094 sset_destroy(&oso.targets);
1097 /* Returns whether a IPFIX row is valid. */
1099 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1101 return ipfix && ipfix->n_targets > 0;
1104 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1106 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1107 const struct bridge *br)
1109 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1112 /* Set IPFIX configuration on 'br'. */
1114 bridge_configure_ipfix(struct bridge *br)
1116 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1117 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1118 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1119 struct ofproto_ipfix_bridge_exporter_options be_opts;
1120 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1121 size_t n_fe_opts = 0;
1123 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1124 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1129 if (!valid_be_cfg && n_fe_opts == 0) {
1130 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1135 memset(&be_opts, 0, sizeof be_opts);
1137 sset_init(&be_opts.targets);
1138 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1140 if (be_cfg->sampling) {
1141 be_opts.sampling_rate = *be_cfg->sampling;
1143 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1145 if (be_cfg->obs_domain_id) {
1146 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1148 if (be_cfg->obs_point_id) {
1149 be_opts.obs_point_id = *be_cfg->obs_point_id;
1151 if (be_cfg->cache_active_timeout) {
1152 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1154 if (be_cfg->cache_max_flows) {
1155 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1159 if (n_fe_opts > 0) {
1160 struct ofproto_ipfix_flow_exporter_options *opts;
1161 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1163 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1164 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1165 opts->collector_set_id = fe_cfg->id;
1166 sset_init(&opts->targets);
1167 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1168 fe_cfg->ipfix->n_targets);
1169 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1170 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1171 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1172 ? *fe_cfg->ipfix->cache_max_flows : 0;
1178 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1182 sset_destroy(&be_opts.targets);
1185 if (n_fe_opts > 0) {
1186 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1188 for (i = 0; i < n_fe_opts; i++) {
1189 sset_destroy(&opts->targets);
1197 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1198 struct ofproto_port_stp_settings *port_s,
1199 int *port_num_counter, unsigned long *port_num_bitmap)
1201 const char *config_str;
1202 struct iface *iface;
1204 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1205 port_s->enable = false;
1208 port_s->enable = true;
1211 /* STP over bonds is not supported. */
1212 if (!list_is_singleton(&port->ifaces)) {
1213 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1215 port_s->enable = false;
1219 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1221 /* Internal ports shouldn't participate in spanning tree, so
1223 if (!strcmp(iface->type, "internal")) {
1224 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1225 port_s->enable = false;
1229 /* STP on mirror output ports is not supported. */
1230 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1231 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1232 port_s->enable = false;
1236 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1238 unsigned long int port_num = strtoul(config_str, NULL, 0);
1239 int port_idx = port_num - 1;
1241 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1242 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1243 port_s->enable = false;
1247 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1248 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1249 port->name, port_num);
1250 port_s->enable = false;
1253 bitmap_set1(port_num_bitmap, port_idx);
1254 port_s->port_num = port_idx;
1256 if (*port_num_counter >= STP_MAX_PORTS) {
1257 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1258 port_s->enable = false;
1262 port_s->port_num = (*port_num_counter)++;
1265 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1267 port_s->path_cost = strtoul(config_str, NULL, 10);
1269 enum netdev_features current;
1272 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1273 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1274 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1277 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1279 port_s->priority = strtoul(config_str, NULL, 0);
1281 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1285 /* Set spanning tree configuration on 'br'. */
1287 bridge_configure_stp(struct bridge *br)
1289 if (!br->cfg->stp_enable) {
1290 ofproto_set_stp(br->ofproto, NULL);
1292 struct ofproto_stp_settings br_s;
1293 const char *config_str;
1295 int port_num_counter;
1296 unsigned long *port_num_bitmap;
1298 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1300 uint8_t ea[ETH_ADDR_LEN];
1302 if (eth_addr_from_string(config_str, ea)) {
1303 br_s.system_id = eth_addr_to_uint64(ea);
1305 br_s.system_id = eth_addr_to_uint64(br->ea);
1306 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1307 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1310 br_s.system_id = eth_addr_to_uint64(br->ea);
1313 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1315 br_s.priority = strtoul(config_str, NULL, 0);
1317 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1320 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1322 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1324 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1327 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1329 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1331 br_s.max_age = STP_DEFAULT_MAX_AGE;
1334 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1336 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1338 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1341 /* Configure STP on the bridge. */
1342 if (ofproto_set_stp(br->ofproto, &br_s)) {
1343 VLOG_ERR("bridge %s: could not enable STP", br->name);
1347 /* Users must either set the port number with the "stp-port-num"
1348 * configuration on all ports or none. If manual configuration
1349 * is not done, then we allocate them sequentially. */
1350 port_num_counter = 0;
1351 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1352 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1353 struct ofproto_port_stp_settings port_s;
1354 struct iface *iface;
1356 port_configure_stp(br->ofproto, port, &port_s,
1357 &port_num_counter, port_num_bitmap);
1359 /* As bonds are not supported, just apply configuration to
1360 * all interfaces. */
1361 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1362 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1364 VLOG_ERR("port %s: could not enable STP", port->name);
1370 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1371 && port_num_counter) {
1372 VLOG_ERR("bridge %s: must manually configure all STP port "
1373 "IDs or none, disabling", br->name);
1374 ofproto_set_stp(br->ofproto, NULL);
1376 bitmap_free(port_num_bitmap);
1381 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1383 const struct port *port = port_lookup(br, name);
1384 return port && port_is_bond_fake_iface(port);
1388 port_is_bond_fake_iface(const struct port *port)
1390 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1394 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1396 struct bridge *br, *next;
1397 struct shash new_br;
1400 /* Collect new bridges' names and types. */
1401 shash_init(&new_br);
1402 for (i = 0; i < cfg->n_bridges; i++) {
1403 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1404 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1406 if (strchr(br_cfg->name, '/')) {
1407 /* Prevent remote ovsdb-server users from accessing arbitrary
1408 * directories, e.g. consider a bridge named "../../../etc/". */
1409 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1411 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1412 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1416 /* Get rid of deleted bridges or those whose types have changed.
1417 * Update 'cfg' of bridges that still exist. */
1418 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1419 br->cfg = shash_find_data(&new_br, br->name);
1420 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1421 br->cfg->datapath_type))) {
1426 /* Add new bridges. */
1427 for (i = 0; i < cfg->n_bridges; i++) {
1428 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1429 struct bridge *br = bridge_lookup(br_cfg->name);
1431 bridge_create(br_cfg);
1435 shash_destroy(&new_br);
1438 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1439 * Returns 0 if successful, otherwise a positive errno value. */
1441 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1442 struct netdev *netdev)
1444 return netdev_set_config(netdev, &iface_cfg->options);
1447 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1448 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1450 * If successful, returns 0 and stores the network device in '*netdevp'. On
1451 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1453 iface_do_create(const struct bridge *br,
1454 const struct ovsrec_interface *iface_cfg,
1455 const struct ovsrec_port *port_cfg,
1456 ofp_port_t *ofp_portp, struct netdev **netdevp)
1458 struct netdev *netdev = NULL;
1461 if (netdev_is_reserved_name(iface_cfg->name)) {
1462 VLOG_WARN("could not create interface %s, name is reserved",
1468 error = netdev_open(iface_cfg->name,
1469 iface_get_type(iface_cfg, br->cfg), &netdev);
1471 VLOG_WARN("could not open network device %s (%s)",
1472 iface_cfg->name, ovs_strerror(error));
1476 error = iface_set_netdev_config(iface_cfg, netdev);
1481 *ofp_portp = iface_pick_ofport(iface_cfg);
1482 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1487 VLOG_INFO("bridge %s: added interface %s on port %d",
1488 br->name, iface_cfg->name, *ofp_portp);
1490 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1491 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1499 netdev_close(netdev);
1503 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1504 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1505 * automatically allocated for the iface. Takes ownership of and
1506 * deallocates 'if_cfg'.
1508 * Return true if an iface is successfully created, false otherwise. */
1510 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1511 const struct ovsrec_port *port_cfg)
1513 struct netdev *netdev;
1514 struct iface *iface;
1515 ofp_port_t ofp_port;
1519 /* Do the bits that can fail up front. */
1520 ovs_assert(!iface_lookup(br, iface_cfg->name));
1521 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev);
1523 iface_clear_db_record(iface_cfg);
1527 /* Get or create the port structure. */
1528 port = port_lookup(br, port_cfg->name);
1530 port = port_create(br, port_cfg);
1533 /* Create the iface structure. */
1534 iface = xzalloc(sizeof *iface);
1535 list_push_back(&port->ifaces, &iface->port_elem);
1536 hmap_insert(&br->iface_by_name, &iface->name_node,
1537 hash_string(iface_cfg->name, 0));
1539 iface->name = xstrdup(iface_cfg->name);
1540 iface->ofp_port = ofp_port;
1541 iface->netdev = netdev;
1542 iface->type = iface_get_type(iface_cfg, br->cfg);
1543 iface->cfg = iface_cfg;
1544 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1545 hash_ofp_port(ofp_port));
1547 /* Populate initial status in database. */
1548 iface_refresh_stats(iface);
1549 iface_refresh_status(iface);
1551 /* Add bond fake iface if necessary. */
1552 if (port_is_bond_fake_iface(port)) {
1553 struct ofproto_port ofproto_port;
1555 if (ofproto_port_query_by_name(br->ofproto, port->name,
1557 struct netdev *netdev;
1560 error = netdev_open(port->name, "internal", &netdev);
1562 ofp_port_t fake_ofp_port = OFPP_NONE;
1563 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1564 netdev_close(netdev);
1566 VLOG_WARN("could not open network device %s (%s)",
1567 port->name, ovs_strerror(error));
1570 /* Already exists, nothing to do. */
1571 ofproto_port_destroy(&ofproto_port);
1578 /* Set forward BPDU option. */
1580 bridge_configure_forward_bpdu(struct bridge *br)
1582 ofproto_set_forward_bpdu(br->ofproto,
1583 smap_get_bool(&br->cfg->other_config,
1588 /* Set MAC learning table configuration for 'br'. */
1590 bridge_configure_mac_table(struct bridge *br)
1592 const char *idle_time_str;
1595 const char *mac_table_size_str;
1598 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1599 idle_time = (idle_time_str && atoi(idle_time_str)
1600 ? atoi(idle_time_str)
1601 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1603 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1604 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1605 ? atoi(mac_table_size_str)
1608 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1612 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1613 const struct port *fake_br, struct iface **hw_addr_iface)
1615 struct hmapx mirror_output_ports;
1617 bool found_addr = false;
1621 /* Mirror output ports don't participate in picking the local hardware
1622 * address. ofproto can't help us find out whether a given port is a
1623 * mirror output because we haven't configured mirrors yet, so we need to
1624 * accumulate them ourselves. */
1625 hmapx_init(&mirror_output_ports);
1626 for (i = 0; i < br->cfg->n_mirrors; i++) {
1627 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1628 if (m->output_port) {
1629 hmapx_add(&mirror_output_ports, m->output_port);
1633 /* Otherwise choose the minimum non-local MAC address among all of the
1635 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1636 uint8_t iface_ea[ETH_ADDR_LEN];
1637 struct iface *candidate;
1638 struct iface *iface;
1640 /* Mirror output ports don't participate. */
1641 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1645 /* Choose the MAC address to represent the port. */
1647 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1648 /* Find the interface with this Ethernet address (if any) so that
1649 * we can provide the correct devname to the caller. */
1650 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1651 uint8_t candidate_ea[ETH_ADDR_LEN];
1652 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1653 && eth_addr_equals(iface_ea, candidate_ea)) {
1658 /* Choose the interface whose MAC address will represent the port.
1659 * The Linux kernel bonding code always chooses the MAC address of
1660 * the first slave added to a bond, and the Fedora networking
1661 * scripts always add slaves to a bond in alphabetical order, so
1662 * for compatibility we choose the interface with the name that is
1663 * first in alphabetical order. */
1664 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1665 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1670 /* The local port doesn't count (since we're trying to choose its
1671 * MAC address anyway). */
1672 if (iface->ofp_port == OFPP_LOCAL) {
1676 /* For fake bridges we only choose from ports with the same tag */
1677 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1678 if (!port->cfg->tag) {
1681 if (*port->cfg->tag != *fake_br->cfg->tag) {
1687 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1693 /* Compare against our current choice. */
1694 if (!eth_addr_is_multicast(iface_ea) &&
1695 !eth_addr_is_local(iface_ea) &&
1696 !eth_addr_is_reserved(iface_ea) &&
1697 !eth_addr_is_zero(iface_ea) &&
1698 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1700 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1701 *hw_addr_iface = iface;
1707 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1708 *hw_addr_iface = NULL;
1711 hmapx_destroy(&mirror_output_ports);
1715 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1716 struct iface **hw_addr_iface)
1719 *hw_addr_iface = NULL;
1721 /* Did the user request a particular MAC? */
1722 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1723 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1724 if (eth_addr_is_multicast(ea)) {
1725 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1726 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1727 } else if (eth_addr_is_zero(ea)) {
1728 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1734 /* Find a local hw address */
1735 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1738 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1739 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1740 * an interface on 'br', then that interface must be passed in as
1741 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1742 * 'hw_addr_iface' must be passed in as a null pointer. */
1744 bridge_pick_datapath_id(struct bridge *br,
1745 const uint8_t bridge_ea[ETH_ADDR_LEN],
1746 struct iface *hw_addr_iface)
1749 * The procedure for choosing a bridge MAC address will, in the most
1750 * ordinary case, also choose a unique MAC that we can use as a datapath
1751 * ID. In some special cases, though, multiple bridges will end up with
1752 * the same MAC address. This is OK for the bridges, but it will confuse
1753 * the OpenFlow controller, because each datapath needs a unique datapath
1756 * Datapath IDs must be unique. It is also very desirable that they be
1757 * stable from one run to the next, so that policy set on a datapath
1760 const char *datapath_id;
1763 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1764 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1768 if (!hw_addr_iface) {
1770 * A purely internal bridge, that is, one that has no non-virtual
1771 * network devices on it at all, is difficult because it has no
1772 * natural unique identifier at all.
1774 * When the host is a XenServer, we handle this case by hashing the
1775 * host's UUID with the name of the bridge. Names of bridges are
1776 * persistent across XenServer reboots, although they can be reused if
1777 * an internal network is destroyed and then a new one is later
1778 * created, so this is fairly effective.
1780 * When the host is not a XenServer, we punt by using a random MAC
1781 * address on each run.
1783 const char *host_uuid = xenserver_get_host_uuid();
1785 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1786 dpid = dpid_from_hash(combined, strlen(combined));
1792 return eth_addr_to_uint64(bridge_ea);
1796 dpid_from_hash(const void *data, size_t n)
1798 uint8_t hash[SHA1_DIGEST_SIZE];
1800 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1801 sha1_bytes(data, n, hash);
1802 eth_addr_mark_random(hash);
1803 return eth_addr_to_uint64(hash);
1807 iface_refresh_status(struct iface *iface)
1811 enum netdev_features current;
1815 uint8_t mac[ETH_ADDR_LEN];
1819 if (iface_is_synthetic(iface)) {
1825 if (!netdev_get_status(iface->netdev, &smap)) {
1826 ovsrec_interface_set_status(iface->cfg, &smap);
1828 ovsrec_interface_set_status(iface->cfg, NULL);
1831 smap_destroy(&smap);
1833 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1834 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1836 ovsrec_interface_set_duplex(iface->cfg,
1837 netdev_features_is_full_duplex(current)
1839 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1841 ovsrec_interface_set_duplex(iface->cfg, NULL);
1842 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1845 error = netdev_get_mtu(iface->netdev, &mtu);
1848 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1850 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1853 error = netdev_get_etheraddr(iface->netdev, mac);
1855 char mac_string[32];
1857 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1858 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1860 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1863 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1864 * if there is no valid ifindex number. */
1865 ifindex64 = netdev_get_ifindex(iface->netdev);
1866 if (ifindex64 < 0) {
1869 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1872 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1875 iface_refresh_cfm_stats(struct iface *iface)
1877 const struct ovsrec_interface *cfg = iface->cfg;
1878 struct ofproto_cfm_status status;
1880 if (!ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1881 iface->ofp_port, &status)) {
1882 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1883 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1884 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1885 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
1886 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1887 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
1889 const char *reasons[CFM_FAULT_N_REASONS];
1890 int64_t cfm_health = status.health;
1891 int64_t cfm_flap_count = status.flap_count;
1892 bool faulted = status.faults != 0;
1895 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
1898 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
1899 int reason = 1 << i;
1900 if (status.faults & reason) {
1901 reasons[j++] = cfm_fault_reason_to_str(reason);
1904 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
1906 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
1908 if (status.remote_opstate >= 0) {
1909 const char *remote_opstate = status.remote_opstate ? "up" : "down";
1910 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
1912 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
1915 ovsrec_interface_set_cfm_remote_mpids(cfg,
1916 (const int64_t *)status.rmps,
1918 if (cfm_health >= 0) {
1919 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
1921 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
1929 iface_refresh_stats(struct iface *iface)
1931 #define IFACE_STATS \
1932 IFACE_STAT(rx_packets, "rx_packets") \
1933 IFACE_STAT(tx_packets, "tx_packets") \
1934 IFACE_STAT(rx_bytes, "rx_bytes") \
1935 IFACE_STAT(tx_bytes, "tx_bytes") \
1936 IFACE_STAT(rx_dropped, "rx_dropped") \
1937 IFACE_STAT(tx_dropped, "tx_dropped") \
1938 IFACE_STAT(rx_errors, "rx_errors") \
1939 IFACE_STAT(tx_errors, "tx_errors") \
1940 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1941 IFACE_STAT(rx_over_errors, "rx_over_err") \
1942 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1943 IFACE_STAT(collisions, "collisions")
1945 #define IFACE_STAT(MEMBER, NAME) + 1
1946 enum { N_IFACE_STATS = IFACE_STATS };
1948 int64_t values[N_IFACE_STATS];
1949 char *keys[N_IFACE_STATS];
1952 struct netdev_stats stats;
1954 if (iface_is_synthetic(iface)) {
1958 /* Intentionally ignore return value, since errors will set 'stats' to
1959 * all-1s, and we will deal with that correctly below. */
1960 netdev_get_stats(iface->netdev, &stats);
1962 /* Copy statistics into keys[] and values[]. */
1964 #define IFACE_STAT(MEMBER, NAME) \
1965 if (stats.MEMBER != UINT64_MAX) { \
1967 values[n] = stats.MEMBER; \
1972 ovs_assert(n <= N_IFACE_STATS);
1974 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
1979 br_refresh_stp_status(struct bridge *br)
1981 struct smap smap = SMAP_INITIALIZER(&smap);
1982 struct ofproto *ofproto = br->ofproto;
1983 struct ofproto_stp_status status;
1985 if (ofproto_get_stp_status(ofproto, &status)) {
1989 if (!status.enabled) {
1990 ovsrec_bridge_set_status(br->cfg, NULL);
1994 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
1995 STP_ID_ARGS(status.bridge_id));
1996 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
1997 STP_ID_ARGS(status.designated_root));
1998 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2000 ovsrec_bridge_set_status(br->cfg, &smap);
2001 smap_destroy(&smap);
2005 port_refresh_stp_status(struct port *port)
2007 struct ofproto *ofproto = port->bridge->ofproto;
2008 struct iface *iface;
2009 struct ofproto_port_stp_status status;
2012 if (port_is_synthetic(port)) {
2016 /* STP doesn't currently support bonds. */
2017 if (!list_is_singleton(&port->ifaces)) {
2018 ovsrec_port_set_status(port->cfg, NULL);
2022 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2023 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2027 if (!status.enabled) {
2028 ovsrec_port_set_status(port->cfg, NULL);
2032 /* Set Status column. */
2034 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2035 smap_add(&smap, "stp_state", stp_state_name(status.state));
2036 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2037 smap_add(&smap, "stp_role", stp_role_name(status.role));
2038 ovsrec_port_set_status(port->cfg, &smap);
2039 smap_destroy(&smap);
2043 port_refresh_stp_stats(struct port *port)
2045 struct ofproto *ofproto = port->bridge->ofproto;
2046 struct iface *iface;
2047 struct ofproto_port_stp_stats stats;
2049 int64_t int_values[3];
2051 if (port_is_synthetic(port)) {
2055 /* STP doesn't currently support bonds. */
2056 if (!list_is_singleton(&port->ifaces)) {
2060 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2061 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2065 if (!stats.enabled) {
2066 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2070 /* Set Statistics column. */
2071 keys[0] = "stp_tx_count";
2072 int_values[0] = stats.tx_count;
2073 keys[1] = "stp_rx_count";
2074 int_values[1] = stats.rx_count;
2075 keys[2] = "stp_error_count";
2076 int_values[2] = stats.error_count;
2078 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2079 ARRAY_SIZE(int_values));
2083 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2085 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2089 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2091 bool enable = enable_system_stats(cfg);
2093 system_stats_enable(enable);
2095 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2100 run_system_stats(void)
2102 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2105 stats = system_stats_run();
2107 struct ovsdb_idl_txn *txn;
2108 struct ovsdb_datum datum;
2110 txn = ovsdb_idl_txn_create(idl);
2111 ovsdb_datum_from_smap(&datum, stats);
2112 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2114 ovsdb_idl_txn_commit(txn);
2115 ovsdb_idl_txn_destroy(txn);
2122 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2125 case OFPCR12_ROLE_EQUAL:
2127 case OFPCR12_ROLE_MASTER:
2129 case OFPCR12_ROLE_SLAVE:
2131 case OFPCR12_ROLE_NOCHANGE:
2133 return "*** INVALID ROLE ***";
2138 refresh_controller_status(void)
2142 const struct ovsrec_controller *cfg;
2146 /* Accumulate status for controllers on all bridges. */
2147 HMAP_FOR_EACH (br, node, &all_bridges) {
2148 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2151 /* Update each controller in the database with current status. */
2152 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2153 struct ofproto_controller_info *cinfo =
2154 shash_find_data(&info, cfg->target);
2157 struct smap smap = SMAP_INITIALIZER(&smap);
2158 const char **values = cinfo->pairs.values;
2159 const char **keys = cinfo->pairs.keys;
2162 for (i = 0; i < cinfo->pairs.n; i++) {
2163 smap_add(&smap, keys[i], values[i]);
2166 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2167 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2169 ovsrec_controller_set_status(cfg, &smap);
2170 smap_destroy(&smap);
2172 ovsrec_controller_set_is_connected(cfg, false);
2173 ovsrec_controller_set_role(cfg, NULL);
2174 ovsrec_controller_set_status(cfg, NULL);
2178 ofproto_free_ofproto_controller_info(&info);
2183 * Some information in the database must be kept as up-to-date as possible to
2184 * allow controllers to respond rapidly to network outages. We call these
2185 * statistics "instant" stats.
2187 * We wish to update these statistics every INSTANT_INTERVAL_MSEC milliseconds,
2188 * assuming that they've changed. The only means we have to determine whether
2189 * they have changed are:
2191 * - Try to commit changes to the database. If nothing changed, then
2192 * ovsdb_idl_txn_commit() returns TXN_UNCHANGED, otherwise some other
2195 * - instant_stats_run() is called late in the run loop, after anything that
2196 * might change any of the instant stats.
2198 * We use these two facts together to avoid waking the process up every
2199 * INSTANT_INTERVAL_MSEC whether there is any change or not.
2202 /* Minimum interval between writing updates to the instant stats to the
2204 #define INSTANT_INTERVAL_MSEC 100
2206 /* Current instant stats database transaction, NULL if there is no ongoing
2208 static struct ovsdb_idl_txn *instant_txn;
2210 /* Next time (in msec on monotonic clock) at which we will update the instant
2212 static long long int instant_next_txn = LLONG_MIN;
2214 /* True if the run loop has run since we last saw that the instant stats were
2215 * unchanged, that is, this is true if we need to wake up at 'instant_next_txn'
2216 * to refresh the instant stats. */
2217 static bool instant_stats_could_have_changed;
2220 instant_stats_run(void)
2222 enum ovsdb_idl_txn_status status;
2224 instant_stats_could_have_changed = true;
2230 if (time_msec() < instant_next_txn) {
2233 instant_next_txn = time_msec() + INSTANT_INTERVAL_MSEC;
2235 seq = seq_read(connectivity_seq_get());
2236 if (seq == connectivity_seqno) {
2239 connectivity_seqno = seq;
2241 instant_txn = ovsdb_idl_txn_create(idl);
2242 HMAP_FOR_EACH (br, node, &all_bridges) {
2243 struct iface *iface;
2246 br_refresh_stp_status(br);
2248 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2249 port_refresh_stp_status(port);
2252 HMAP_FOR_EACH (iface, name_node, &br->iface_by_name) {
2253 enum netdev_flags flags;
2255 const char *link_state;
2256 int64_t link_resets;
2259 if (iface_is_synthetic(iface)) {
2263 current = ofproto_port_is_lacp_current(br->ofproto,
2267 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2269 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2272 error = netdev_get_flags(iface->netdev, &flags);
2274 const char *state = flags & NETDEV_UP ? "up" : "down";
2275 ovsrec_interface_set_admin_state(iface->cfg, state);
2277 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2280 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2281 ovsrec_interface_set_link_state(iface->cfg, link_state);
2283 link_resets = netdev_get_carrier_resets(iface->netdev);
2284 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2286 iface_refresh_cfm_stats(iface);
2289 ofproto_port_get_bfd_status(br->ofproto, iface->ofp_port,
2291 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2292 smap_destroy(&smap);
2297 status = ovsdb_idl_txn_commit(instant_txn);
2298 if (status != TXN_INCOMPLETE) {
2299 ovsdb_idl_txn_destroy(instant_txn);
2302 if (status == TXN_UNCHANGED) {
2303 instant_stats_could_have_changed = false;
2308 instant_stats_wait(void)
2311 ovsdb_idl_txn_wait(instant_txn);
2312 } else if (instant_stats_could_have_changed) {
2313 poll_timer_wait_until(instant_next_txn);
2324 /* Let each datapath type do the work that it needs to do. */
2326 ofproto_enumerate_types(&types);
2327 SSET_FOR_EACH (type, &types) {
2328 ofproto_type_run(type);
2330 sset_destroy(&types);
2332 /* Let each bridge do the work that it needs to do. */
2333 HMAP_FOR_EACH (br, node, &all_bridges) {
2334 ofproto_run(br->ofproto);
2341 static struct ovsrec_open_vswitch null_cfg;
2342 const struct ovsrec_open_vswitch *cfg;
2344 bool vlan_splinters_changed;
2347 ovsrec_open_vswitch_init(&null_cfg);
2351 if (ovsdb_idl_is_lock_contended(idl)) {
2352 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2353 struct bridge *br, *next_br;
2355 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2356 "disabling this process (pid %ld) until it goes away",
2357 (long int) getpid());
2359 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2362 /* Since we will not be running system_stats_run() in this process
2363 * with the current situation of multiple ovs-vswitchd daemons,
2364 * disable system stats collection. */
2365 system_stats_enable(false);
2367 } else if (!ovsdb_idl_has_lock(idl)) {
2370 cfg = ovsrec_open_vswitch_first(idl);
2372 /* Initialize the ofproto library. This only needs to run once, but
2373 * it must be done after the configuration is set. If the
2374 * initialization has already occurred, bridge_init_ofproto()
2375 * returns immediately. */
2376 bridge_init_ofproto(cfg);
2378 /* Once the value of flow-restore-wait is false, we no longer should
2379 * check its value from the database. */
2380 if (cfg && ofproto_get_flow_restore_wait()) {
2381 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2382 "flow-restore-wait", false));
2387 /* Re-configure SSL. We do this on every trip through the main loop,
2388 * instead of just when the database changes, because the contents of the
2389 * key and certificate files can change without the database changing.
2391 * We do this before bridge_reconfigure() because that function might
2392 * initiate SSL connections and thus requires SSL to be configured. */
2393 if (cfg && cfg->ssl) {
2394 const struct ovsrec_ssl *ssl = cfg->ssl;
2396 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2397 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2400 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2401 * usage has changed. */
2402 vlan_splinters_changed = false;
2403 if (vlan_splinters_enabled_anywhere) {
2404 HMAP_FOR_EACH (br, node, &all_bridges) {
2405 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2406 vlan_splinters_changed = true;
2412 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2413 struct ovsdb_idl_txn *txn;
2415 idl_seqno = ovsdb_idl_get_seqno(idl);
2416 txn = ovsdb_idl_txn_create(idl);
2417 bridge_reconfigure(cfg ? cfg : &null_cfg);
2420 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2423 /* If we are completing our initial configuration for this run
2424 * of ovs-vswitchd, then keep the transaction around to monitor
2425 * it for completion. */
2426 if (initial_config_done) {
2427 ovsdb_idl_txn_commit(txn);
2428 ovsdb_idl_txn_destroy(txn);
2430 initial_config_done = true;
2431 daemonize_txn = txn;
2435 if (daemonize_txn) {
2436 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2437 if (status != TXN_INCOMPLETE) {
2438 ovsdb_idl_txn_destroy(daemonize_txn);
2439 daemonize_txn = NULL;
2441 /* ovs-vswitchd has completed initialization, so allow the
2442 * process that forked us to exit successfully. */
2443 daemonize_complete();
2445 vlog_enable_async();
2447 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2451 /* Refresh interface and mirror stats if necessary. */
2452 if (time_msec() >= iface_stats_timer) {
2454 struct ovsdb_idl_txn *txn;
2456 txn = ovsdb_idl_txn_create(idl);
2457 HMAP_FOR_EACH (br, node, &all_bridges) {
2461 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2462 struct iface *iface;
2464 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2465 iface_refresh_stats(iface);
2466 iface_refresh_status(iface);
2469 port_refresh_stp_stats(port);
2472 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2473 mirror_refresh_stats(m);
2477 refresh_controller_status();
2478 ovsdb_idl_txn_commit(txn);
2479 ovsdb_idl_txn_destroy(txn); /* XXX */
2482 iface_stats_timer = time_msec() + IFACE_STATS_INTERVAL;
2486 instant_stats_run();
2495 ovsdb_idl_wait(idl);
2496 if (daemonize_txn) {
2497 ovsdb_idl_txn_wait(daemonize_txn);
2501 ofproto_enumerate_types(&types);
2502 SSET_FOR_EACH (type, &types) {
2503 ofproto_type_wait(type);
2505 sset_destroy(&types);
2507 if (!hmap_is_empty(&all_bridges)) {
2510 HMAP_FOR_EACH (br, node, &all_bridges) {
2511 ofproto_wait(br->ofproto);
2513 poll_timer_wait_until(iface_stats_timer);
2516 system_stats_wait();
2517 instant_stats_wait();
2520 /* Adds some memory usage statistics for bridges into 'usage', for use with
2521 * memory_report(). */
2523 bridge_get_memory_usage(struct simap *usage)
2530 ofproto_enumerate_types(&types);
2531 SSET_FOR_EACH (type, &types) {
2532 ofproto_type_get_memory_usage(type, usage);
2534 sset_destroy(&types);
2536 HMAP_FOR_EACH (br, node, &all_bridges) {
2537 ofproto_get_memory_usage(br->ofproto, usage);
2541 /* QoS unixctl user interface functions. */
2543 struct qos_unixctl_show_cbdata {
2545 struct iface *iface;
2549 qos_unixctl_show_queue(unsigned int queue_id,
2550 const struct smap *details,
2551 struct iface *iface,
2554 struct netdev_queue_stats stats;
2555 struct smap_node *node;
2558 ds_put_cstr(ds, "\n");
2560 ds_put_format(ds, "Queue %u:\n", queue_id);
2562 ds_put_cstr(ds, "Default:\n");
2565 SMAP_FOR_EACH (node, details) {
2566 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2569 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2571 if (stats.tx_packets != UINT64_MAX) {
2572 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2575 if (stats.tx_bytes != UINT64_MAX) {
2576 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2579 if (stats.tx_errors != UINT64_MAX) {
2580 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2583 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2584 queue_id, ovs_strerror(error));
2589 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2590 const char *argv[], void *aux OVS_UNUSED)
2592 struct ds ds = DS_EMPTY_INITIALIZER;
2593 struct smap smap = SMAP_INITIALIZER(&smap);
2594 struct iface *iface;
2596 struct smap_node *node;
2598 iface = iface_find(argv[1]);
2600 unixctl_command_reply_error(conn, "no such interface");
2604 netdev_get_qos(iface->netdev, &type, &smap);
2606 if (*type != '\0') {
2607 struct netdev_queue_dump dump;
2608 struct smap details;
2609 unsigned int queue_id;
2611 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2613 SMAP_FOR_EACH (node, &smap) {
2614 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2617 smap_init(&details);
2618 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2619 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2621 smap_destroy(&details);
2623 unixctl_command_reply(conn, ds_cstr(&ds));
2625 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2626 unixctl_command_reply_error(conn, ds_cstr(&ds));
2629 smap_destroy(&smap);
2633 /* Bridge reconfiguration functions. */
2635 bridge_create(const struct ovsrec_bridge *br_cfg)
2639 ovs_assert(!bridge_lookup(br_cfg->name));
2640 br = xzalloc(sizeof *br);
2642 br->name = xstrdup(br_cfg->name);
2643 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2646 /* Derive the default Ethernet address from the bridge's UUID. This should
2647 * be unique and it will be stable between ovs-vswitchd runs. */
2648 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2649 eth_addr_mark_random(br->default_ea);
2651 hmap_init(&br->ports);
2652 hmap_init(&br->ifaces);
2653 hmap_init(&br->iface_by_name);
2654 hmap_init(&br->mirrors);
2656 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2660 bridge_destroy(struct bridge *br)
2663 struct mirror *mirror, *next_mirror;
2664 struct port *port, *next_port;
2666 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2669 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2670 mirror_destroy(mirror);
2673 hmap_remove(&all_bridges, &br->node);
2674 ofproto_destroy(br->ofproto);
2675 hmap_destroy(&br->ifaces);
2676 hmap_destroy(&br->ports);
2677 hmap_destroy(&br->iface_by_name);
2678 hmap_destroy(&br->mirrors);
2685 static struct bridge *
2686 bridge_lookup(const char *name)
2690 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2691 if (!strcmp(br->name, name)) {
2698 /* Handle requests for a listing of all flows known by the OpenFlow
2699 * stack, including those normally hidden. */
2701 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2702 const char *argv[], void *aux OVS_UNUSED)
2707 br = bridge_lookup(argv[1]);
2709 unixctl_command_reply_error(conn, "Unknown bridge");
2714 ofproto_get_all_flows(br->ofproto, &results);
2716 unixctl_command_reply(conn, ds_cstr(&results));
2717 ds_destroy(&results);
2720 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2721 * connections and reconnect. If BRIDGE is not specified, then all bridges
2722 * drop their controller connections and reconnect. */
2724 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2725 const char *argv[], void *aux OVS_UNUSED)
2729 br = bridge_lookup(argv[1]);
2731 unixctl_command_reply_error(conn, "Unknown bridge");
2734 ofproto_reconnect_controllers(br->ofproto);
2736 HMAP_FOR_EACH (br, node, &all_bridges) {
2737 ofproto_reconnect_controllers(br->ofproto);
2740 unixctl_command_reply(conn, NULL);
2744 bridge_get_controllers(const struct bridge *br,
2745 struct ovsrec_controller ***controllersp)
2747 struct ovsrec_controller **controllers;
2748 size_t n_controllers;
2750 controllers = br->cfg->controller;
2751 n_controllers = br->cfg->n_controller;
2753 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2759 *controllersp = controllers;
2761 return n_controllers;
2765 bridge_collect_wanted_ports(struct bridge *br,
2766 const unsigned long int *splinter_vlans,
2767 struct shash *wanted_ports)
2771 shash_init(wanted_ports);
2773 for (i = 0; i < br->cfg->n_ports; i++) {
2774 const char *name = br->cfg->ports[i]->name;
2775 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2776 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2780 if (bridge_get_controllers(br, NULL)
2781 && !shash_find(wanted_ports, br->name)) {
2782 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2783 br->name, br->name);
2785 ovsrec_interface_init(&br->synth_local_iface);
2786 ovsrec_port_init(&br->synth_local_port);
2788 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2789 br->synth_local_port.n_interfaces = 1;
2790 br->synth_local_port.name = br->name;
2792 br->synth_local_iface.name = br->name;
2793 br->synth_local_iface.type = "internal";
2795 br->synth_local_ifacep = &br->synth_local_iface;
2797 shash_add(wanted_ports, br->name, &br->synth_local_port);
2800 if (splinter_vlans) {
2801 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2805 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2806 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2807 * 'br' needs to complete its configuration. */
2809 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2811 struct shash_node *port_node;
2812 struct port *port, *next;
2814 /* Get rid of deleted ports.
2815 * Get rid of deleted interfaces on ports that still exist. */
2816 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2817 port->cfg = shash_find_data(wanted_ports, port->name);
2821 port_del_ifaces(port);
2825 /* Update iface->cfg and iface->type in interfaces that still exist. */
2826 SHASH_FOR_EACH (port_node, wanted_ports) {
2827 const struct ovsrec_port *port = port_node->data;
2830 for (i = 0; i < port->n_interfaces; i++) {
2831 const struct ovsrec_interface *cfg = port->interfaces[i];
2832 struct iface *iface = iface_lookup(br, cfg->name);
2833 const char *type = iface_get_type(cfg, br->cfg);
2838 } else if (!strcmp(type, "null")) {
2839 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2840 " may be removed in February 2013. Please email"
2841 " dev@openvswitch.org with concerns.",
2844 /* We will add new interfaces later. */
2850 /* Initializes 'oc' appropriately as a management service controller for
2853 * The caller must free oc->target when it is no longer needed. */
2855 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2856 struct ofproto_controller *oc)
2858 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2859 oc->max_backoff = 0;
2860 oc->probe_interval = 60;
2861 oc->band = OFPROTO_OUT_OF_BAND;
2863 oc->burst_limit = 0;
2864 oc->enable_async_msgs = true;
2867 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2869 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2870 struct ofproto_controller *oc)
2874 oc->target = c->target;
2875 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2876 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2877 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2878 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2879 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2880 oc->burst_limit = (c->controller_burst_limit
2881 ? *c->controller_burst_limit : 0);
2882 oc->enable_async_msgs = (!c->enable_async_messages
2883 || *c->enable_async_messages);
2884 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2885 if (dscp < 0 || dscp > 63) {
2886 dscp = DSCP_DEFAULT;
2891 /* Configures the IP stack for 'br''s local interface properly according to the
2892 * configuration in 'c'. */
2894 bridge_configure_local_iface_netdev(struct bridge *br,
2895 struct ovsrec_controller *c)
2897 struct netdev *netdev;
2898 struct in_addr mask, gateway;
2900 struct iface *local_iface;
2903 /* If there's no local interface or no IP address, give up. */
2904 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2905 if (!local_iface || !c->local_ip
2906 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2910 /* Bring up the local interface. */
2911 netdev = local_iface->netdev;
2912 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2914 /* Configure the IP address and netmask. */
2915 if (!c->local_netmask
2916 || !inet_pton(AF_INET, c->local_netmask, &mask)
2918 mask.s_addr = guess_netmask(ip.s_addr);
2920 if (!netdev_set_in4(netdev, ip, mask)) {
2921 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2922 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2925 /* Configure the default gateway. */
2926 if (c->local_gateway
2927 && inet_pton(AF_INET, c->local_gateway, &gateway)
2928 && gateway.s_addr) {
2929 if (!netdev_add_router(netdev, gateway)) {
2930 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2931 br->name, IP_ARGS(gateway.s_addr));
2936 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2937 * in either string are treated as equal to any number of slashes in the other,
2938 * e.g. "x///y" is equal to "x/y".
2940 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2941 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2942 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2943 * 'b' against a prefix of 'a'.
2946 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
2948 const char *b_start = b;
2950 if (b - b_start >= b_stoplen) {
2952 } else if (*a != *b) {
2954 } else if (*a == '/') {
2955 a += strspn(a, "/");
2956 b += strspn(b, "/");
2957 } else if (*a == '\0') {
2967 bridge_configure_remotes(struct bridge *br,
2968 const struct sockaddr_in *managers, size_t n_managers)
2970 bool disable_in_band;
2972 struct ovsrec_controller **controllers;
2973 size_t n_controllers;
2975 enum ofproto_fail_mode fail_mode;
2977 struct ofproto_controller *ocs;
2981 /* Check if we should disable in-band control on this bridge. */
2982 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
2985 /* Set OpenFlow queue ID for in-band control. */
2986 ofproto_set_in_band_queue(br->ofproto,
2987 smap_get_int(&br->cfg->other_config,
2988 "in-band-queue", -1));
2990 if (disable_in_band) {
2991 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
2993 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
2996 n_controllers = bridge_get_controllers(br, &controllers);
2998 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3001 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3002 for (i = 0; i < n_controllers; i++) {
3003 struct ovsrec_controller *c = controllers[i];
3005 if (!strncmp(c->target, "punix:", 6)
3006 || !strncmp(c->target, "unix:", 5)) {
3007 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3010 if (!strncmp(c->target, "unix:", 5)) {
3011 /* Connect to a listening socket */
3012 whitelist = xasprintf("unix:%s/", ovs_rundir());
3013 if (strchr(c->target, '/') &&
3014 !equal_pathnames(c->target, whitelist,
3015 strlen(whitelist))) {
3016 /* Absolute path specified, but not in ovs_rundir */
3017 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3018 "controller \"%s\" due to possibility for "
3019 "remote exploit. Instead, specify socket "
3020 "in whitelisted \"%s\" or connect to "
3021 "\"unix:%s/%s.mgmt\" (which is always "
3022 "available without special configuration).",
3023 br->name, c->target, whitelist,
3024 ovs_rundir(), br->name);
3029 whitelist = xasprintf("punix:%s/%s.controller",
3030 ovs_rundir(), br->name);
3031 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3032 /* Prevent remote ovsdb-server users from accessing
3033 * arbitrary Unix domain sockets and overwriting arbitrary
3035 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3036 "controller \"%s\" due to possibility of "
3037 "overwriting local files. Instead, specify "
3038 "whitelisted \"%s\" or connect to "
3039 "\"unix:%s/%s.mgmt\" (which is always "
3040 "available without special configuration).",
3041 br->name, c->target, whitelist,
3042 ovs_rundir(), br->name);
3051 bridge_configure_local_iface_netdev(br, c);
3052 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3053 if (disable_in_band) {
3054 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3059 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3060 bridge_get_allowed_versions(br));
3061 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3064 /* Set the fail-mode. */
3065 fail_mode = !br->cfg->fail_mode
3066 || !strcmp(br->cfg->fail_mode, "standalone")
3067 ? OFPROTO_FAIL_STANDALONE
3068 : OFPROTO_FAIL_SECURE;
3069 ofproto_set_fail_mode(br->ofproto, fail_mode);
3071 /* Configure OpenFlow controller connection snooping. */
3072 if (!ofproto_has_snoops(br->ofproto)) {
3076 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3077 ovs_rundir(), br->name));
3078 ofproto_set_snoops(br->ofproto, &snoops);
3079 sset_destroy(&snoops);
3084 bridge_configure_tables(struct bridge *br)
3086 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3090 n_tables = ofproto_get_n_tables(br->ofproto);
3092 for (i = 0; i < n_tables; i++) {
3093 struct ofproto_table_settings s;
3096 s.max_flows = UINT_MAX;
3099 s.n_prefix_fields = 0;
3100 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3102 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3103 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3106 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3107 s.max_flows = *cfg->flow_limit;
3109 if (cfg->overflow_policy
3110 && !strcmp(cfg->overflow_policy, "evict")) {
3112 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3113 for (k = 0; k < cfg->n_groups; k++) {
3114 const char *string = cfg->groups[k];
3117 msg = mf_parse_subfield__(&s.groups[k], &string);
3119 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3120 "'groups' (%s)", br->name, i, msg);
3122 } else if (*string) {
3123 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3124 "element '%s' contains trailing garbage",
3125 br->name, i, cfg->groups[k]);
3131 /* Prefix lookup fields. */
3132 s.n_prefix_fields = 0;
3133 for (k = 0; k < cfg->n_prefixes; k++) {
3134 const char *name = cfg->prefixes[k];
3135 const struct mf_field *mf = mf_from_name(name);
3137 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3141 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3142 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3143 "%s", br->name, name);
3146 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3147 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3148 "field not used: %s", br->name, name);
3151 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3153 if (s.n_prefix_fields > 0) {
3155 struct ds ds = DS_EMPTY_INITIALIZER;
3156 for (k = 0; k < s.n_prefix_fields; k++) {
3158 ds_put_char(&ds, ',');
3160 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3162 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3163 br->name, i, ds_cstr(&ds));
3168 ofproto_configure_table(br->ofproto, i, &s);
3172 for (; j < br->cfg->n_flow_tables; j++) {
3173 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3174 "%"PRId64" not supported by this datapath", br->name,
3175 br->cfg->key_flow_tables[j]);
3180 bridge_configure_dp_desc(struct bridge *br)
3182 ofproto_set_dp_desc(br->ofproto,
3183 smap_get(&br->cfg->other_config, "dp-desc"));
3186 /* Port functions. */
3188 static struct port *
3189 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3193 port = xzalloc(sizeof *port);
3195 port->name = xstrdup(cfg->name);
3197 list_init(&port->ifaces);
3199 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3203 /* Deletes interfaces from 'port' that are no longer configured for it. */
3205 port_del_ifaces(struct port *port)
3207 struct iface *iface, *next;
3208 struct sset new_ifaces;
3211 /* Collect list of new interfaces. */
3212 sset_init(&new_ifaces);
3213 for (i = 0; i < port->cfg->n_interfaces; i++) {
3214 const char *name = port->cfg->interfaces[i]->name;
3215 const char *type = port->cfg->interfaces[i]->type;
3216 if (strcmp(type, "null")) {
3217 sset_add(&new_ifaces, name);
3221 /* Get rid of deleted interfaces. */
3222 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3223 if (!sset_contains(&new_ifaces, iface->name)) {
3224 iface_destroy(iface);
3228 sset_destroy(&new_ifaces);
3232 port_destroy(struct port *port)
3235 struct bridge *br = port->bridge;
3236 struct iface *iface, *next;
3239 ofproto_bundle_unregister(br->ofproto, port);
3242 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3243 iface_destroy__(iface);
3246 hmap_remove(&br->ports, &port->hmap_node);
3252 static struct port *
3253 port_lookup(const struct bridge *br, const char *name)
3257 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3259 if (!strcmp(port->name, name)) {
3267 enable_lacp(struct port *port, bool *activep)
3269 if (!port->cfg->lacp) {
3270 /* XXX when LACP implementation has been sufficiently tested, enable by
3271 * default and make active on bonded ports. */
3273 } else if (!strcmp(port->cfg->lacp, "off")) {
3275 } else if (!strcmp(port->cfg->lacp, "active")) {
3278 } else if (!strcmp(port->cfg->lacp, "passive")) {
3282 VLOG_WARN("port %s: unknown LACP mode %s",
3283 port->name, port->cfg->lacp);
3288 static struct lacp_settings *
3289 port_configure_lacp(struct port *port, struct lacp_settings *s)
3291 const char *lacp_time, *system_id;
3294 if (!enable_lacp(port, &s->active)) {
3298 s->name = port->name;
3300 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3302 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3303 ETH_ADDR_SCAN_ARGS(s->id))) {
3304 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3305 " address.", port->name, system_id);
3309 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3312 if (eth_addr_is_zero(s->id)) {
3313 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3317 /* Prefer bondable links if unspecified. */
3318 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3320 s->priority = (priority > 0 && priority <= UINT16_MAX
3322 : UINT16_MAX - !list_is_short(&port->ifaces));
3324 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3325 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3327 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3328 "lacp-fallback-ab", false);
3334 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3336 int priority, portid, key;
3338 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3339 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3341 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3343 if (portid <= 0 || portid > UINT16_MAX) {
3344 portid = ofp_to_u16(iface->ofp_port);
3347 if (priority <= 0 || priority > UINT16_MAX) {
3348 priority = UINT16_MAX;
3351 if (key < 0 || key > UINT16_MAX) {
3355 s->name = iface->name;
3357 s->priority = priority;
3362 port_configure_bond(struct port *port, struct bond_settings *s)
3364 const char *detect_s;
3365 struct iface *iface;
3366 int miimon_interval;
3368 s->name = port->name;
3370 if (port->cfg->bond_mode) {
3371 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3372 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3373 port->name, port->cfg->bond_mode,
3374 bond_mode_to_string(s->balance));
3377 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3379 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3380 * active-backup. At some point we should remove this warning. */
3381 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3382 " in previous versions, the default bond_mode was"
3383 " balance-slb", port->name,
3384 bond_mode_to_string(s->balance));
3386 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3387 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3388 "please use another bond type or disable flood_vlans",
3392 miimon_interval = smap_get_int(&port->cfg->other_config,
3393 "bond-miimon-interval", 0);
3394 if (miimon_interval <= 0) {
3395 miimon_interval = 200;
3398 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3399 if (!detect_s || !strcmp(detect_s, "carrier")) {
3400 miimon_interval = 0;
3401 } else if (strcmp(detect_s, "miimon")) {
3402 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3403 "defaulting to carrier", port->name, detect_s);
3404 miimon_interval = 0;
3407 s->up_delay = MAX(0, port->cfg->bond_updelay);
3408 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3409 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3410 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3411 "bond-rebalance-interval", 10000);
3412 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3413 s->rebalance_interval = 1000;
3416 s->fake_iface = port->cfg->bond_fake_iface;
3418 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3419 "lacp-fallback-ab", false);
3421 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3422 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3426 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3427 * instead of obtaining it from the database. */
3429 port_is_synthetic(const struct port *port)
3431 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3434 /* Interface functions. */
3437 iface_is_internal(const struct ovsrec_interface *iface,
3438 const struct ovsrec_bridge *br)
3440 /* The local port and "internal" ports are always "internal". */
3441 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3444 /* Returns the correct network device type for interface 'iface' in bridge
3447 iface_get_type(const struct ovsrec_interface *iface,
3448 const struct ovsrec_bridge *br)
3452 /* The local port always has type "internal". Other ports take
3453 * their type from the database and default to "system" if none is
3455 if (iface_is_internal(iface, br)) {
3458 type = iface->type[0] ? iface->type : "system";
3461 return ofproto_port_open_type(br->datapath_type, type);
3465 iface_destroy__(struct iface *iface)
3468 struct port *port = iface->port;
3469 struct bridge *br = port->bridge;
3471 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3472 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3475 if (iface->ofp_port != OFPP_NONE) {
3476 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3479 list_remove(&iface->port_elem);
3480 hmap_remove(&br->iface_by_name, &iface->name_node);
3482 netdev_close(iface->netdev);
3490 iface_destroy(struct iface *iface)
3493 struct port *port = iface->port;
3495 iface_destroy__(iface);
3496 if (list_is_empty(&port->ifaces)) {
3502 static struct iface *
3503 iface_lookup(const struct bridge *br, const char *name)
3505 struct iface *iface;
3507 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3508 &br->iface_by_name) {
3509 if (!strcmp(iface->name, name)) {
3517 static struct iface *
3518 iface_find(const char *name)
3520 const struct bridge *br;
3522 HMAP_FOR_EACH (br, node, &all_bridges) {
3523 struct iface *iface = iface_lookup(br, name);
3532 static struct iface *
3533 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3535 struct iface *iface;
3537 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3539 if (iface->ofp_port == ofp_port) {
3546 /* Set Ethernet address of 'iface', if one is specified in the configuration
3549 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3551 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3552 struct iface *hw_addr_iface;
3554 if (strcmp(iface->type, "internal")) {
3558 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3560 } else if (port->cfg->fake_bridge) {
3561 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3562 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3567 if (iface->ofp_port == OFPP_LOCAL) {
3568 VLOG_ERR("interface %s: ignoring mac in Interface record "
3569 "(use Bridge record to set local port's mac)",
3571 } else if (eth_addr_is_multicast(mac)) {
3572 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3575 int error = netdev_set_etheraddr(iface->netdev, mac);
3577 VLOG_ERR("interface %s: setting MAC failed (%s)",
3578 iface->name, ovs_strerror(error));
3584 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3586 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3588 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3589 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3590 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3594 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3595 * sets the "ofport" field to -1.
3597 * This is appropriate when 'if_cfg''s interface cannot be created or is
3598 * otherwise invalid. */
3600 iface_clear_db_record(const struct ovsrec_interface *if_cfg)
3602 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3603 iface_set_ofport(if_cfg, OFPP_NONE);
3604 ovsrec_interface_set_status(if_cfg, NULL);
3605 ovsrec_interface_set_admin_state(if_cfg, NULL);
3606 ovsrec_interface_set_duplex(if_cfg, NULL);
3607 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3608 ovsrec_interface_set_link_state(if_cfg, NULL);
3609 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3610 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3611 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3612 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3613 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3614 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3615 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3616 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3621 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3623 union ovsdb_atom atom;
3625 atom.integer = target;
3626 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3630 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3632 struct ofpbuf queues_buf;
3634 ofpbuf_init(&queues_buf, 0);
3636 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3637 netdev_set_qos(iface->netdev, NULL, NULL);
3639 const struct ovsdb_datum *queues;
3640 struct netdev_queue_dump dump;
3641 unsigned int queue_id;
3642 struct smap details;
3646 /* Configure top-level Qos for 'iface'. */
3647 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3649 /* Deconfigure queues that were deleted. */
3650 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3652 smap_init(&details);
3653 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3654 if (!queue_ids_include(queues, queue_id)) {
3655 netdev_delete_queue(iface->netdev, queue_id);
3658 smap_destroy(&details);
3660 /* Configure queues for 'iface'. */
3662 for (i = 0; i < qos->n_queues; i++) {
3663 const struct ovsrec_queue *queue = qos->value_queues[i];
3664 unsigned int queue_id = qos->key_queues[i];
3666 if (queue_id == 0) {
3670 if (queue->n_dscp == 1) {
3671 struct ofproto_port_queue *port_queue;
3673 port_queue = ofpbuf_put_uninit(&queues_buf,
3674 sizeof *port_queue);
3675 port_queue->queue = queue_id;
3676 port_queue->dscp = queue->dscp[0];
3679 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3682 struct smap details;
3684 smap_init(&details);
3685 netdev_set_queue(iface->netdev, 0, &details);
3686 smap_destroy(&details);
3690 if (iface->ofp_port != OFPP_NONE) {
3691 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3692 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3694 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3695 port_queues, n_queues);
3698 netdev_set_policing(iface->netdev,
3699 iface->cfg->ingress_policing_rate,
3700 iface->cfg->ingress_policing_burst);
3702 ofpbuf_uninit(&queues_buf);
3706 iface_configure_cfm(struct iface *iface)
3708 const struct ovsrec_interface *cfg = iface->cfg;
3709 const char *opstate_str;
3710 const char *cfm_ccm_vlan;
3711 struct cfm_settings s;
3712 struct smap netdev_args;
3714 if (!cfg->n_cfm_mpid) {
3715 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3719 s.check_tnl_key = false;
3720 smap_init(&netdev_args);
3721 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3722 const char *key = smap_get(&netdev_args, "key");
3723 const char *in_key = smap_get(&netdev_args, "in_key");
3725 s.check_tnl_key = (key && !strcmp(key, "flow"))
3726 || (in_key && !strcmp(in_key, "flow"));
3728 smap_destroy(&netdev_args);
3730 s.mpid = *cfg->cfm_mpid;
3731 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3732 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3733 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3735 if (s.interval <= 0) {
3739 if (!cfm_ccm_vlan) {
3741 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3742 s.ccm_vlan = CFM_RANDOM_VLAN;
3744 s.ccm_vlan = atoi(cfm_ccm_vlan);
3745 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3750 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3752 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3754 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3755 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3757 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3760 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3761 * instead of obtaining it from the database. */
3763 iface_is_synthetic(const struct iface *iface)
3765 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3769 iface_validate_ofport__(size_t n, int64_t *ofport)
3771 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3772 ? u16_to_ofp(*ofport)
3777 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3779 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3783 iface_pick_ofport(const struct ovsrec_interface *cfg)
3785 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3786 return (requested_ofport != OFPP_NONE
3788 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3791 /* Port mirroring. */
3793 static struct mirror *
3794 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3798 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3799 if (uuid_equals(uuid, &m->uuid)) {
3807 bridge_configure_mirrors(struct bridge *br)
3809 const struct ovsdb_datum *mc;
3810 unsigned long *flood_vlans;
3811 struct mirror *m, *next;
3814 /* Get rid of deleted mirrors. */
3815 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3816 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3817 union ovsdb_atom atom;
3819 atom.uuid = m->uuid;
3820 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3825 /* Add new mirrors and reconfigure existing ones. */
3826 for (i = 0; i < br->cfg->n_mirrors; i++) {
3827 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3828 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3830 m = mirror_create(br, cfg);
3833 if (!mirror_configure(m)) {
3838 /* Update flooded vlans (for RSPAN). */
3839 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3840 br->cfg->n_flood_vlans);
3841 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3842 bitmap_free(flood_vlans);
3845 static struct mirror *
3846 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3850 m = xzalloc(sizeof *m);
3851 m->uuid = cfg->header_.uuid;
3852 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3854 m->name = xstrdup(cfg->name);
3860 mirror_destroy(struct mirror *m)
3863 struct bridge *br = m->bridge;
3866 ofproto_mirror_unregister(br->ofproto, m);
3869 hmap_remove(&br->mirrors, &m->hmap_node);
3876 mirror_collect_ports(struct mirror *m,
3877 struct ovsrec_port **in_ports, int n_in_ports,
3878 void ***out_portsp, size_t *n_out_portsp)
3880 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3881 size_t n_out_ports = 0;
3884 for (i = 0; i < n_in_ports; i++) {
3885 const char *name = in_ports[i]->name;
3886 struct port *port = port_lookup(m->bridge, name);
3888 out_ports[n_out_ports++] = port;
3890 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3891 "port %s", m->bridge->name, m->name, name);
3894 *out_portsp = out_ports;
3895 *n_out_portsp = n_out_ports;
3899 mirror_configure(struct mirror *m)
3901 const struct ovsrec_mirror *cfg = m->cfg;
3902 struct ofproto_mirror_settings s;
3905 if (strcmp(cfg->name, m->name)) {
3907 m->name = xstrdup(cfg->name);
3911 /* Get output port or VLAN. */
3912 if (cfg->output_port) {
3913 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3914 if (!s.out_bundle) {
3915 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3916 m->bridge->name, m->name);
3919 s.out_vlan = UINT16_MAX;
3921 if (cfg->output_vlan) {
3922 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3923 "output vlan; ignoring output vlan",
3924 m->bridge->name, m->name);
3926 } else if (cfg->output_vlan) {
3927 /* The database should prevent invalid VLAN values. */
3928 s.out_bundle = NULL;
3929 s.out_vlan = *cfg->output_vlan;
3931 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
3932 m->bridge->name, m->name);
3936 /* Get port selection. */
3937 if (cfg->select_all) {
3938 size_t n_ports = hmap_count(&m->bridge->ports);
3939 void **ports = xmalloc(n_ports * sizeof *ports);
3944 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
3954 /* Get ports, dropping ports that don't exist.
3955 * The IDL ensures that there are no duplicates. */
3956 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
3957 &s.srcs, &s.n_srcs);
3958 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
3959 &s.dsts, &s.n_dsts);
3962 /* Get VLAN selection. */
3963 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
3966 ofproto_mirror_register(m->bridge->ofproto, m, &s);
3969 if (s.srcs != s.dsts) {
3978 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
3980 * This is deprecated. It is only for compatibility with broken device drivers
3981 * in old versions of Linux that do not properly support VLANs when VLAN
3982 * devices are not used. When broken device drivers are no longer in
3983 * widespread use, we will delete these interfaces. */
3985 static struct ovsrec_port **recs;
3986 static size_t n_recs, allocated_recs;
3988 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
3989 * splinters are reconfigured. */
3991 register_rec(struct ovsrec_port *rec)
3993 if (n_recs >= allocated_recs) {
3994 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
3996 recs[n_recs++] = rec;
3999 /* Frees all of the ports registered with register_reg(). */
4001 free_registered_recs(void)
4005 for (i = 0; i < n_recs; i++) {
4006 struct ovsrec_port *port = recs[i];
4009 for (j = 0; j < port->n_interfaces; j++) {
4010 struct ovsrec_interface *iface = port->interfaces[j];
4015 smap_destroy(&port->other_config);
4016 free(port->interfaces);
4024 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4027 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4029 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4033 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4036 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4037 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4038 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4041 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4042 * use, returns NULL.
4044 * Updates 'vlan_splinters_enabled_anywhere'. */
4045 static unsigned long int *
4046 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4048 unsigned long int *splinter_vlans;
4049 struct sset splinter_ifaces;
4050 const char *real_dev_name;
4051 struct shash *real_devs;
4052 struct shash_node *node;
4056 /* Free space allocated for synthesized ports and interfaces, since we're
4057 * in the process of reconstructing all of them. */
4058 free_registered_recs();
4060 splinter_vlans = bitmap_allocate(4096);
4061 sset_init(&splinter_ifaces);
4062 vlan_splinters_enabled_anywhere = false;
4063 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4064 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4067 for (j = 0; j < br_cfg->n_ports; j++) {
4068 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4071 for (k = 0; k < port_cfg->n_interfaces; k++) {
4072 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4074 if (vlan_splinters_is_enabled(iface_cfg)) {
4075 vlan_splinters_enabled_anywhere = true;
4076 sset_add(&splinter_ifaces, iface_cfg->name);
4077 vlan_bitmap_from_array__(port_cfg->trunks,
4083 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4084 bitmap_set1(splinter_vlans, *port_cfg->tag);
4089 if (!vlan_splinters_enabled_anywhere) {
4090 free(splinter_vlans);
4091 sset_destroy(&splinter_ifaces);
4095 HMAP_FOR_EACH (br, node, &all_bridges) {
4097 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4101 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4102 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4103 * device to be created for it. */
4104 bitmap_set0(splinter_vlans, 0);
4105 bitmap_set0(splinter_vlans, 4095);
4107 /* Delete all VLAN devices that we don't need. */
4109 real_devs = vlandev_get_real_devs();
4110 SHASH_FOR_EACH (node, real_devs) {
4111 const struct vlan_real_dev *real_dev = node->data;
4112 const struct vlan_dev *vlan_dev;
4113 bool real_dev_has_splinters;
4115 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4117 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4118 if (!real_dev_has_splinters
4119 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4120 struct netdev *netdev;
4122 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4123 if (!netdev_get_in4(netdev, NULL, NULL) ||
4124 !netdev_get_in6(netdev, NULL)) {
4125 /* It has an IP address configured, so we don't own
4126 * it. Don't delete it. */
4128 vlandev_del(vlan_dev->name);
4130 netdev_close(netdev);
4137 /* Add all VLAN devices that we need. */
4138 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4141 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4142 if (!vlandev_get_name(real_dev_name, vid)) {
4143 vlandev_add(real_dev_name, vid);
4150 sset_destroy(&splinter_ifaces);
4152 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4153 free(splinter_vlans);
4156 return splinter_vlans;
4159 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4162 configure_splinter_port(struct port *port)
4164 struct ofproto *ofproto = port->bridge->ofproto;
4165 ofp_port_t realdev_ofp_port;
4166 const char *realdev_name;
4167 struct iface *vlandev, *realdev;
4169 ofproto_bundle_unregister(port->bridge->ofproto, port);
4171 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4174 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4175 realdev = iface_lookup(port->bridge, realdev_name);
4176 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4178 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4182 static struct ovsrec_port *
4183 synthesize_splinter_port(const char *real_dev_name,
4184 const char *vlan_dev_name, int vid)
4186 struct ovsrec_interface *iface;
4187 struct ovsrec_port *port;
4189 iface = xmalloc(sizeof *iface);
4190 ovsrec_interface_init(iface);
4191 iface->name = xstrdup(vlan_dev_name);
4192 iface->type = "system";
4194 port = xmalloc(sizeof *port);
4195 ovsrec_port_init(port);
4196 port->interfaces = xmemdup(&iface, sizeof iface);
4197 port->n_interfaces = 1;
4198 port->name = xstrdup(vlan_dev_name);
4199 port->vlan_mode = "splinter";
4200 port->tag = xmalloc(sizeof *port->tag);
4203 smap_add(&port->other_config, "realdev", real_dev_name);
4209 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4210 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4211 * 1-bit in the 'splinter_vlans' bitmap. */
4213 add_vlan_splinter_ports(struct bridge *br,
4214 const unsigned long int *splinter_vlans,
4215 struct shash *ports)
4219 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4220 * we're modifying 'ports'. */
4221 for (i = 0; i < br->cfg->n_ports; i++) {
4222 const char *name = br->cfg->ports[i]->name;
4223 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4226 for (j = 0; j < port_cfg->n_interfaces; j++) {
4227 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4229 if (vlan_splinters_is_enabled(iface_cfg)) {
4230 const char *real_dev_name;
4233 real_dev_name = iface_cfg->name;
4234 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4235 const char *vlan_dev_name;
4237 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4239 && !shash_find(ports, vlan_dev_name)) {
4240 shash_add(ports, vlan_dev_name,
4241 synthesize_splinter_port(
4242 real_dev_name, vlan_dev_name, vid));
4251 mirror_refresh_stats(struct mirror *m)
4253 struct ofproto *ofproto = m->bridge->ofproto;
4254 uint64_t tx_packets, tx_bytes;
4257 size_t stat_cnt = 0;
4259 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4260 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4264 if (tx_packets != UINT64_MAX) {
4265 keys[stat_cnt] = "tx_packets";
4266 values[stat_cnt] = tx_packets;
4269 if (tx_bytes != UINT64_MAX) {
4270 keys[stat_cnt] = "tx_bytes";
4271 values[stat_cnt] = tx_bytes;
4275 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);