1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
21 #include "async-append.h"
25 #include "connectivity.h"
29 #include "dynamic-string.h"
36 #include "mac-learning.h"
37 #include "mcast-snooping.h"
38 #include "meta-flow.h"
40 #include "ofp-print.h"
43 #include "ofproto/bond.h"
44 #include "ofproto/ofproto.h"
45 #include "poll-loop.h"
50 #include "socket-util.h"
52 #include "stream-ssl.h"
54 #include "system-stats.h"
59 #include "lib/vswitch-idl.h"
60 #include "xenserver.h"
62 #include "sflow_api.h"
63 #include "vlan-bitmap.h"
65 VLOG_DEFINE_THIS_MODULE(bridge);
67 COVERAGE_DEFINE(bridge_reconfigure);
70 /* These members are always valid.
72 * They are immutable: they never change between iface_create() and
74 struct list port_elem; /* Element in struct port's "ifaces" list. */
75 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
76 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
77 struct port *port; /* Containing port. */
78 char *name; /* Host network device name. */
79 struct netdev *netdev; /* Network device. */
80 ofp_port_t ofp_port; /* OpenFlow port number. */
83 /* These members are valid only within bridge_reconfigure(). */
84 const char *type; /* Usually same as cfg->type. */
85 const struct ovsrec_interface *cfg;
89 struct uuid uuid; /* UUID of this "mirror" record in database. */
90 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
91 struct bridge *bridge;
93 const struct ovsrec_mirror *cfg;
97 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
98 struct bridge *bridge;
101 const struct ovsrec_port *cfg;
103 /* An ordinary bridge port has 1 interface.
104 * A bridge port for bonding has at least 2 interfaces. */
105 struct list ifaces; /* List of "struct iface"s. */
109 struct hmap_node node; /* In 'all_bridges'. */
110 char *name; /* User-specified arbitrary name. */
111 char *type; /* Datapath type. */
112 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
113 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
114 const struct ovsrec_bridge *cfg;
116 /* OpenFlow switch processing. */
117 struct ofproto *ofproto; /* OpenFlow switch. */
120 struct hmap ports; /* "struct port"s indexed by name. */
121 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
122 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
124 /* Port mirroring. */
125 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
127 /* Used during reconfiguration. */
128 struct shash wanted_ports;
130 /* Synthetic local port if necessary. */
131 struct ovsrec_port synth_local_port;
132 struct ovsrec_interface synth_local_iface;
133 struct ovsrec_interface *synth_local_ifacep;
136 /* All bridges, indexed by name. */
137 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
139 /* OVSDB IDL used to obtain configuration. */
140 static struct ovsdb_idl *idl;
142 /* We want to complete daemonization, fully detaching from our parent process,
143 * only after we have completed our initial configuration, committed our state
144 * to the database, and received confirmation back from the database server
145 * that it applied the commit. This allows our parent process to know that,
146 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
147 * to have already been filled in. (It is only "very likely" rather than
148 * certain because there is always a slim possibility that the transaction will
149 * fail or that some other client has added new bridges, ports, etc. while
150 * ovs-vswitchd was configuring using an old configuration.)
152 * We only need to do this once for our initial configuration at startup, so
153 * 'initial_config_done' tracks whether we've already done it. While we are
154 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
155 * itself and is otherwise NULL. */
156 static bool initial_config_done;
157 static struct ovsdb_idl_txn *daemonize_txn;
159 /* Most recently processed IDL sequence number. */
160 static unsigned int idl_seqno;
162 /* Track changes to port connectivity. */
163 static uint64_t connectivity_seqno = LLONG_MIN;
165 /* Status update to database.
167 * Some information in the database must be kept as up-to-date as possible to
168 * allow controllers to respond rapidly to network outages. Those status are
169 * updated via the 'status_txn'.
171 * We use the global connectivity sequence number to detect the status change.
172 * Also, to prevent the status update from sending too much to the database,
173 * we check the return status of each update transaction and do not start new
174 * update if the previous transaction status is 'TXN_INCOMPLETE'.
176 * 'statux_txn' is NULL if there is no ongoing status update.
178 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
179 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
180 * which will cause the main thread wake up soon and retry the status update.
182 static struct ovsdb_idl_txn *status_txn;
183 static bool status_txn_try_again;
185 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
186 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
187 #define STATUS_CHECK_AGAIN_MSEC 100
189 /* Each time this timer expires, the bridge fetches interface and mirror
190 * statistics and pushes them into the database. */
191 static int stats_timer_interval;
192 static long long int stats_timer = LLONG_MIN;
194 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
195 * expensive. This can cause bridge_reconfigure() to take a long time during
196 * which no other work can be done. To deal with this problem, we limit port
197 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
198 * call to bridge_reconfigure(). If there is more work to do after the limit
199 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
200 * This allows the rest of the code to catch up on important things like
201 * forwarding packets. */
202 #define OFP_PORT_ACTION_WINDOW 10
204 static void add_del_bridges(const struct ovsrec_open_vswitch *);
205 static void bridge_run__(void);
206 static void bridge_create(const struct ovsrec_bridge *);
207 static void bridge_destroy(struct bridge *);
208 static struct bridge *bridge_lookup(const char *name);
209 static unixctl_cb_func bridge_unixctl_dump_flows;
210 static unixctl_cb_func bridge_unixctl_reconnect;
211 static size_t bridge_get_controllers(const struct bridge *br,
212 struct ovsrec_controller ***controllersp);
213 static void bridge_collect_wanted_ports(struct bridge *,
214 const unsigned long *splinter_vlans,
215 struct shash *wanted_ports);
216 static void bridge_delete_ofprotos(void);
217 static void bridge_delete_or_reconfigure_ports(struct bridge *);
218 static void bridge_del_ports(struct bridge *,
219 const struct shash *wanted_ports);
220 static void bridge_add_ports(struct bridge *,
221 const struct shash *wanted_ports);
223 static void bridge_configure_datapath_id(struct bridge *);
224 static void bridge_configure_netflow(struct bridge *);
225 static void bridge_configure_forward_bpdu(struct bridge *);
226 static void bridge_configure_mac_table(struct bridge *);
227 static void bridge_configure_mcast_snooping(struct bridge *);
228 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
229 static void bridge_configure_ipfix(struct bridge *);
230 static void bridge_configure_stp(struct bridge *);
231 static void bridge_configure_tables(struct bridge *);
232 static void bridge_configure_dp_desc(struct bridge *);
233 static void bridge_configure_remotes(struct bridge *,
234 const struct sockaddr_in *managers,
236 static void bridge_pick_local_hw_addr(struct bridge *,
237 uint8_t ea[ETH_ADDR_LEN],
238 struct iface **hw_addr_iface);
239 static uint64_t bridge_pick_datapath_id(struct bridge *,
240 const uint8_t bridge_ea[ETH_ADDR_LEN],
241 struct iface *hw_addr_iface);
242 static uint64_t dpid_from_hash(const void *, size_t nbytes);
243 static bool bridge_has_bond_fake_iface(const struct bridge *,
245 static bool port_is_bond_fake_iface(const struct port *);
247 static unixctl_cb_func qos_unixctl_show;
249 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
250 static void port_del_ifaces(struct port *);
251 static void port_destroy(struct port *);
252 static struct port *port_lookup(const struct bridge *, const char *name);
253 static void port_configure(struct port *);
254 static struct lacp_settings *port_configure_lacp(struct port *,
255 struct lacp_settings *);
256 static void port_configure_bond(struct port *, struct bond_settings *);
257 static bool port_is_synthetic(const struct port *);
259 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
260 static void run_system_stats(void);
262 static void bridge_configure_mirrors(struct bridge *);
263 static struct mirror *mirror_create(struct bridge *,
264 const struct ovsrec_mirror *);
265 static void mirror_destroy(struct mirror *);
266 static bool mirror_configure(struct mirror *);
267 static void mirror_refresh_stats(struct mirror *);
269 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
270 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
271 const struct ovsrec_port *);
272 static bool iface_is_internal(const struct ovsrec_interface *iface,
273 const struct ovsrec_bridge *br);
274 static const char *iface_get_type(const struct ovsrec_interface *,
275 const struct ovsrec_bridge *);
276 static void iface_destroy(struct iface *);
277 static void iface_destroy__(struct iface *);
278 static struct iface *iface_lookup(const struct bridge *, const char *name);
279 static struct iface *iface_find(const char *name);
280 static struct iface *iface_from_ofp_port(const struct bridge *,
281 ofp_port_t ofp_port);
282 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
283 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
284 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
285 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
286 static void iface_configure_cfm(struct iface *);
287 static void iface_refresh_cfm_stats(struct iface *);
288 static void iface_refresh_stats(struct iface *);
289 static void iface_refresh_netdev_status(struct iface *);
290 static void iface_refresh_ofproto_status(struct iface *);
291 static bool iface_is_synthetic(const struct iface *);
292 static ofp_port_t iface_get_requested_ofp_port(
293 const struct ovsrec_interface *);
294 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
296 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
298 * This is deprecated. It is only for compatibility with broken device drivers
299 * in old versions of Linux that do not properly support VLANs when VLAN
300 * devices are not used. When broken device drivers are no longer in
301 * widespread use, we will delete these interfaces. */
303 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
304 static bool vlan_splinters_enabled_anywhere;
306 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
307 static unsigned long int *collect_splinter_vlans(
308 const struct ovsrec_open_vswitch *);
309 static void configure_splinter_port(struct port *);
310 static void add_vlan_splinter_ports(struct bridge *,
311 const unsigned long int *splinter_vlans,
312 struct shash *ports);
315 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
317 struct shash iface_hints;
318 static bool initialized = false;
325 shash_init(&iface_hints);
328 for (i = 0; i < cfg->n_bridges; i++) {
329 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
332 for (j = 0; j < br_cfg->n_ports; j++) {
333 struct ovsrec_port *port_cfg = br_cfg->ports[j];
336 for (k = 0; k < port_cfg->n_interfaces; k++) {
337 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
338 struct iface_hint *iface_hint;
340 iface_hint = xmalloc(sizeof *iface_hint);
341 iface_hint->br_name = br_cfg->name;
342 iface_hint->br_type = br_cfg->datapath_type;
343 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
345 shash_add(&iface_hints, if_cfg->name, iface_hint);
351 ofproto_init(&iface_hints);
353 shash_destroy_free_data(&iface_hints);
357 /* Public functions. */
359 /* Initializes the bridge module, configuring it to obtain its configuration
360 * from an OVSDB server accessed over 'remote', which should be a string in a
361 * form acceptable to ovsdb_idl_create(). */
363 bridge_init(const char *remote)
365 /* Create connection to database. */
366 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
367 idl_seqno = ovsdb_idl_get_seqno(idl);
368 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
369 ovsdb_idl_verify_write_only(idl);
371 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
372 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
373 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
374 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
375 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
376 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
377 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
379 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
380 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
381 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
383 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
384 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
385 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
387 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
388 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
389 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
390 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
391 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
392 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
393 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
394 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
395 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
396 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
397 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
398 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
399 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
400 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
401 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
402 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
403 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
404 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
405 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
406 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
407 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
409 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
410 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
411 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
412 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
414 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
416 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
418 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
419 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
421 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
422 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
423 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
424 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
426 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
427 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
428 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
429 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
430 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
432 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
434 /* Register unixctl commands. */
435 unixctl_command_register("qos/show", "interface", 1, 1,
436 qos_unixctl_show, NULL);
437 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
438 bridge_unixctl_dump_flows, NULL);
439 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
440 bridge_unixctl_reconnect, NULL);
450 struct bridge *br, *next_br;
452 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
455 ovsdb_idl_destroy(idl);
458 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
459 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
460 * responsible for freeing '*managersp' (with free()).
462 * You may be asking yourself "why does ovs-vswitchd care?", because
463 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
464 * should not be and in fact is not directly involved in that. But
465 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
466 * it has to tell in-band control where the managers are to enable that.
467 * (Thus, only managers connected in-band are collected.)
470 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
471 struct sockaddr_in **managersp, size_t *n_managersp)
473 struct sockaddr_in *managers = NULL;
474 size_t n_managers = 0;
478 /* Collect all of the potential targets from the "targets" columns of the
479 * rows pointed to by "manager_options", excluding any that are
482 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
483 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
485 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
486 sset_find_and_delete(&targets, m->target);
488 sset_add(&targets, m->target);
492 /* Now extract the targets' IP addresses. */
493 if (!sset_is_empty(&targets)) {
496 managers = xmalloc(sset_count(&targets) * sizeof *managers);
497 SSET_FOR_EACH (target, &targets) {
498 struct sockaddr_storage ss;
500 if (stream_parse_target_with_default_port(target, OVSDB_OLD_PORT,
502 && ss.ss_family == AF_INET) {
503 managers[n_managers++] = *(struct sockaddr_in *) &ss;
507 sset_destroy(&targets);
509 *managersp = managers;
510 *n_managersp = n_managers;
514 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
516 unsigned long int *splinter_vlans;
517 struct sockaddr_in *managers;
518 struct bridge *br, *next;
519 int sflow_bridge_number;
522 COVERAGE_INC(bridge_reconfigure);
524 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
525 OFPROTO_FLOW_LIMIT_DEFAULT));
526 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
527 OFPROTO_MAX_IDLE_DEFAULT));
530 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
531 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
533 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
534 * to 'ovs_cfg', with only very minimal configuration otherwise.
536 * This is mostly an update to bridge data structures. Nothing is pushed
537 * down to ofproto or lower layers. */
538 add_del_bridges(ovs_cfg);
539 splinter_vlans = collect_splinter_vlans(ovs_cfg);
540 HMAP_FOR_EACH (br, node, &all_bridges) {
541 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
542 bridge_del_ports(br, &br->wanted_ports);
544 free(splinter_vlans);
546 /* Start pushing configuration changes down to the ofproto layer:
548 * - Delete ofprotos that are no longer configured.
550 * - Delete ports that are no longer configured.
552 * - Reconfigure existing ports to their desired configurations, or
553 * delete them if not possible.
555 * We have to do all the deletions before we can do any additions, because
556 * the ports to be added might require resources that will be freed up by
557 * deletions (they might especially overlap in name). */
558 bridge_delete_ofprotos();
559 HMAP_FOR_EACH (br, node, &all_bridges) {
561 bridge_delete_or_reconfigure_ports(br);
565 /* Finish pushing configuration changes to the ofproto layer:
567 * - Create ofprotos that are missing.
569 * - Add ports that are missing. */
570 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
574 error = ofproto_create(br->name, br->type, &br->ofproto);
576 VLOG_ERR("failed to create bridge %s: %s", br->name,
577 ovs_strerror(error));
578 shash_destroy(&br->wanted_ports);
583 HMAP_FOR_EACH (br, node, &all_bridges) {
584 bridge_add_ports(br, &br->wanted_ports);
585 shash_destroy(&br->wanted_ports);
588 reconfigure_system_stats(ovs_cfg);
590 /* Complete the configuration. */
591 sflow_bridge_number = 0;
592 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
593 HMAP_FOR_EACH (br, node, &all_bridges) {
596 /* We need the datapath ID early to allow LACP ports to use it as the
597 * default system ID. */
598 bridge_configure_datapath_id(br);
600 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
603 port_configure(port);
605 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
606 iface_set_ofport(iface->cfg, iface->ofp_port);
607 /* Clear eventual previous errors */
608 ovsrec_interface_set_error(iface->cfg, NULL);
609 iface_configure_cfm(iface);
610 iface_configure_qos(iface, port->cfg->qos);
611 iface_set_mac(br, port, iface);
612 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
616 bridge_configure_mirrors(br);
617 bridge_configure_forward_bpdu(br);
618 bridge_configure_mac_table(br);
619 bridge_configure_mcast_snooping(br);
620 bridge_configure_remotes(br, managers, n_managers);
621 bridge_configure_netflow(br);
622 bridge_configure_sflow(br, &sflow_bridge_number);
623 bridge_configure_ipfix(br);
624 bridge_configure_stp(br);
625 bridge_configure_tables(br);
626 bridge_configure_dp_desc(br);
630 /* The ofproto-dpif provider does some final reconfiguration in its
631 * ->type_run() function. We have to call it before notifying the database
632 * client that reconfiguration is complete, otherwise there is a very
633 * narrow race window in which e.g. ofproto/trace will not recognize the
634 * new configuration (sometimes this causes unit test failures). */
638 /* Delete ofprotos which aren't configured or have the wrong type. Create
639 * ofprotos which don't exist but need to. */
641 bridge_delete_ofprotos(void)
648 /* Delete ofprotos with no bridge or with the wrong type. */
651 ofproto_enumerate_types(&types);
652 SSET_FOR_EACH (type, &types) {
655 ofproto_enumerate_names(type, &names);
656 SSET_FOR_EACH (name, &names) {
657 br = bridge_lookup(name);
658 if (!br || strcmp(type, br->type)) {
659 ofproto_delete(name, type);
663 sset_destroy(&names);
664 sset_destroy(&types);
668 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
670 if (*n >= *allocated) {
671 ports = x2nrealloc(ports, allocated, sizeof *ports);
673 ports[(*n)++] = port;
678 bridge_delete_or_reconfigure_ports(struct bridge *br)
680 struct ofproto_port ofproto_port;
681 struct ofproto_port_dump dump;
683 struct sset ofproto_ports;
684 struct port *port, *port_next;
686 /* List of "ofp_port"s to delete. We make a list instead of deleting them
687 * right away because ofproto implementations aren't necessarily able to
688 * iterate through a changing list of ports in an entirely robust way. */
695 sset_init(&ofproto_ports);
697 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
698 * that are not configured in the database. (This commonly happens when
699 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
701 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
702 * that have the wrong OpenFlow port number (and arrange to add them back
703 * with the correct OpenFlow port number). */
704 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
705 ofp_port_t requested_ofp_port;
708 sset_add(&ofproto_ports, ofproto_port.name);
710 iface = iface_lookup(br, ofproto_port.name);
712 /* No such iface is configured, so we should delete this
715 * As a corner case exception, keep the port if it's a bond fake
717 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
718 && !strcmp(ofproto_port.type, "internal")) {
724 if (strcmp(ofproto_port.type, iface->type)
725 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
726 /* The interface is the wrong type or can't be configured.
731 /* If the requested OpenFlow port for 'iface' changed, and it's not
732 * already the correct port, then we might want to temporarily delete
733 * this interface, so we can add it back again with the new OpenFlow
735 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
736 if (iface->ofp_port != OFPP_LOCAL &&
737 requested_ofp_port != OFPP_NONE &&
738 requested_ofp_port != iface->ofp_port) {
739 ofp_port_t victim_request;
740 struct iface *victim;
742 /* Check for an existing OpenFlow port currently occupying
743 * 'iface''s requested port number. If there isn't one, then
744 * delete this port. Otherwise we need to consider further. */
745 victim = iface_from_ofp_port(br, requested_ofp_port);
750 /* 'victim' is a port currently using 'iface''s requested port
751 * number. Unless 'victim' specifically requested that port
752 * number, too, then we can delete both 'iface' and 'victim'
753 * temporarily. (We'll add both of them back again later with new
754 * OpenFlow port numbers.)
756 * If 'victim' did request port number 'requested_ofp_port', just
757 * like 'iface', then that's a configuration inconsistency that we
758 * can't resolve. We might as well let it keep its current port
760 victim_request = iface_get_requested_ofp_port(victim->cfg);
761 if (victim_request != requested_ofp_port) {
762 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
763 iface_destroy(victim);
772 iface_destroy(iface);
773 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
775 for (i = 0; i < n; i++) {
776 ofproto_port_del(br->ofproto, del[i]);
780 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
781 * that we didn't see when we iterated through the datapath, i.e. ports
782 * that disappeared underneath use. This is an unusual situation, but it
783 * can happen in some cases:
785 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
786 * idea but could happen).
788 * - The port represented a device that disappeared, e.g. a tuntap
789 * device destroyed via "tunctl -d", a physical Ethernet device
790 * whose module was just unloaded via "rmmod", or a virtual NIC for a
791 * VM whose VM was just terminated. */
792 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
793 struct iface *iface, *iface_next;
795 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
796 if (!sset_contains(&ofproto_ports, iface->name)) {
797 iface_destroy__(iface);
801 if (list_is_empty(&port->ifaces)) {
805 sset_destroy(&ofproto_ports);
809 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
810 bool with_requested_port)
812 struct shash_node *port_node;
814 SHASH_FOR_EACH (port_node, wanted_ports) {
815 const struct ovsrec_port *port_cfg = port_node->data;
818 for (i = 0; i < port_cfg->n_interfaces; i++) {
819 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
820 ofp_port_t requested_ofp_port;
822 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
823 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
824 struct iface *iface = iface_lookup(br, iface_cfg->name);
827 iface_create(br, iface_cfg, port_cfg);
835 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
837 /* First add interfaces that request a particular port number. */
838 bridge_add_ports__(br, wanted_ports, true);
840 /* Then add interfaces that want automatic port number assignment.
841 * We add these afterward to avoid accidentally taking a specifically
842 * requested port number. */
843 bridge_add_ports__(br, wanted_ports, false);
847 port_configure(struct port *port)
849 const struct ovsrec_port *cfg = port->cfg;
850 struct bond_settings bond_settings;
851 struct lacp_settings lacp_settings;
852 struct ofproto_bundle_settings s;
855 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
856 configure_splinter_port(port);
865 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
866 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
867 s.slaves[s.n_slaves++] = iface->ofp_port;
872 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
876 /* Get VLAN trunks. */
879 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
883 if (cfg->vlan_mode) {
884 if (!strcmp(cfg->vlan_mode, "access")) {
885 s.vlan_mode = PORT_VLAN_ACCESS;
886 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
887 s.vlan_mode = PORT_VLAN_TRUNK;
888 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
889 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
890 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
891 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
893 /* This "can't happen" because ovsdb-server should prevent it. */
894 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
895 "back to trunk mode", port->name, cfg->vlan_mode);
896 s.vlan_mode = PORT_VLAN_TRUNK;
900 s.vlan_mode = PORT_VLAN_ACCESS;
902 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
906 s.vlan_mode = PORT_VLAN_TRUNK;
909 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
912 /* Get LACP settings. */
913 s.lacp = port_configure_lacp(port, &lacp_settings);
917 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
918 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
919 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
922 s.lacp_slaves = NULL;
925 /* Get bond settings. */
926 if (s.n_slaves > 1) {
927 s.bond = &bond_settings;
928 port_configure_bond(port, &bond_settings);
931 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
932 netdev_set_miimon_interval(iface->netdev, 0);
937 ofproto_bundle_register(port->bridge->ofproto, port, &s);
945 /* Pick local port hardware address and datapath ID for 'br'. */
947 bridge_configure_datapath_id(struct bridge *br)
949 uint8_t ea[ETH_ADDR_LEN];
951 struct iface *local_iface;
952 struct iface *hw_addr_iface;
955 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
956 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
958 int error = netdev_set_etheraddr(local_iface->netdev, ea);
960 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
961 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
962 "Ethernet address: %s",
963 br->name, ovs_strerror(error));
966 memcpy(br->ea, ea, ETH_ADDR_LEN);
968 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
969 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
970 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
971 ofproto_set_datapath_id(br->ofproto, dpid);
974 dpid_string = xasprintf("%016"PRIx64, dpid);
975 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
979 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
982 bridge_get_allowed_versions(struct bridge *br)
984 if (!br->cfg->n_protocols)
987 return ofputil_versions_from_strings(br->cfg->protocols,
988 br->cfg->n_protocols);
991 /* Set NetFlow configuration on 'br'. */
993 bridge_configure_netflow(struct bridge *br)
995 struct ovsrec_netflow *cfg = br->cfg->netflow;
996 struct netflow_options opts;
999 ofproto_set_netflow(br->ofproto, NULL);
1003 memset(&opts, 0, sizeof opts);
1005 /* Get default NetFlow configuration from datapath.
1006 * Apply overrides from 'cfg'. */
1007 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1008 if (cfg->engine_type) {
1009 opts.engine_type = *cfg->engine_type;
1011 if (cfg->engine_id) {
1012 opts.engine_id = *cfg->engine_id;
1015 /* Configure active timeout interval. */
1016 opts.active_timeout = cfg->active_timeout;
1017 if (!opts.active_timeout) {
1018 opts.active_timeout = -1;
1019 } else if (opts.active_timeout < 0) {
1020 VLOG_WARN("bridge %s: active timeout interval set to negative "
1021 "value, using default instead (%d seconds)", br->name,
1022 NF_ACTIVE_TIMEOUT_DEFAULT);
1023 opts.active_timeout = -1;
1026 /* Add engine ID to interface number to disambiguate bridgs? */
1027 opts.add_id_to_iface = cfg->add_id_to_interface;
1028 if (opts.add_id_to_iface) {
1029 if (opts.engine_id > 0x7f) {
1030 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1031 "another vswitch, choose an engine id less than 128",
1034 if (hmap_count(&br->ports) > 508) {
1035 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1036 "another port when more than 508 ports are used",
1042 sset_init(&opts.collectors);
1043 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1046 if (ofproto_set_netflow(br->ofproto, &opts)) {
1047 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1049 sset_destroy(&opts.collectors);
1052 /* Set sFlow configuration on 'br'. */
1054 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1056 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1057 struct ovsrec_controller **controllers;
1058 struct ofproto_sflow_options oso;
1059 size_t n_controllers;
1063 ofproto_set_sflow(br->ofproto, NULL);
1067 memset(&oso, 0, sizeof oso);
1069 sset_init(&oso.targets);
1070 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1072 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1073 if (cfg->sampling) {
1074 oso.sampling_rate = *cfg->sampling;
1077 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1079 oso.polling_interval = *cfg->polling;
1082 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1084 oso.header_len = *cfg->header;
1087 oso.sub_id = (*sflow_bridge_number)++;
1088 oso.agent_device = cfg->agent;
1090 oso.control_ip = NULL;
1091 n_controllers = bridge_get_controllers(br, &controllers);
1092 for (i = 0; i < n_controllers; i++) {
1093 if (controllers[i]->local_ip) {
1094 oso.control_ip = controllers[i]->local_ip;
1098 ofproto_set_sflow(br->ofproto, &oso);
1100 sset_destroy(&oso.targets);
1103 /* Returns whether a IPFIX row is valid. */
1105 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1107 return ipfix && ipfix->n_targets > 0;
1110 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1112 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1113 const struct bridge *br)
1115 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1118 /* Set IPFIX configuration on 'br'. */
1120 bridge_configure_ipfix(struct bridge *br)
1122 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1123 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1124 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1125 struct ofproto_ipfix_bridge_exporter_options be_opts;
1126 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1127 size_t n_fe_opts = 0;
1129 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1130 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1135 if (!valid_be_cfg && n_fe_opts == 0) {
1136 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1141 memset(&be_opts, 0, sizeof be_opts);
1143 sset_init(&be_opts.targets);
1144 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1146 if (be_cfg->sampling) {
1147 be_opts.sampling_rate = *be_cfg->sampling;
1149 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1151 if (be_cfg->obs_domain_id) {
1152 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1154 if (be_cfg->obs_point_id) {
1155 be_opts.obs_point_id = *be_cfg->obs_point_id;
1157 if (be_cfg->cache_active_timeout) {
1158 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1160 if (be_cfg->cache_max_flows) {
1161 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1165 if (n_fe_opts > 0) {
1166 struct ofproto_ipfix_flow_exporter_options *opts;
1167 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1169 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1170 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1171 opts->collector_set_id = fe_cfg->id;
1172 sset_init(&opts->targets);
1173 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1174 fe_cfg->ipfix->n_targets);
1175 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1176 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1177 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1178 ? *fe_cfg->ipfix->cache_max_flows : 0;
1184 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1188 sset_destroy(&be_opts.targets);
1191 if (n_fe_opts > 0) {
1192 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1194 for (i = 0; i < n_fe_opts; i++) {
1195 sset_destroy(&opts->targets);
1203 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1204 struct ofproto_port_stp_settings *port_s,
1205 int *port_num_counter, unsigned long *port_num_bitmap)
1207 const char *config_str;
1208 struct iface *iface;
1210 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1211 port_s->enable = false;
1214 port_s->enable = true;
1217 /* STP over bonds is not supported. */
1218 if (!list_is_singleton(&port->ifaces)) {
1219 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1221 port_s->enable = false;
1225 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1227 /* Internal ports shouldn't participate in spanning tree, so
1229 if (!strcmp(iface->type, "internal")) {
1230 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1231 port_s->enable = false;
1235 /* STP on mirror output ports is not supported. */
1236 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1237 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1238 port_s->enable = false;
1242 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1244 unsigned long int port_num = strtoul(config_str, NULL, 0);
1245 int port_idx = port_num - 1;
1247 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1248 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1249 port_s->enable = false;
1253 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1254 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1255 port->name, port_num);
1256 port_s->enable = false;
1259 bitmap_set1(port_num_bitmap, port_idx);
1260 port_s->port_num = port_idx;
1262 if (*port_num_counter >= STP_MAX_PORTS) {
1263 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1264 port_s->enable = false;
1268 port_s->port_num = (*port_num_counter)++;
1271 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1273 port_s->path_cost = strtoul(config_str, NULL, 10);
1275 enum netdev_features current;
1278 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1279 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1280 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1283 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1285 port_s->priority = strtoul(config_str, NULL, 0);
1287 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1291 /* Set spanning tree configuration on 'br'. */
1293 bridge_configure_stp(struct bridge *br)
1295 if (!br->cfg->stp_enable) {
1296 ofproto_set_stp(br->ofproto, NULL);
1298 struct ofproto_stp_settings br_s;
1299 const char *config_str;
1301 int port_num_counter;
1302 unsigned long *port_num_bitmap;
1304 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1306 uint8_t ea[ETH_ADDR_LEN];
1308 if (eth_addr_from_string(config_str, ea)) {
1309 br_s.system_id = eth_addr_to_uint64(ea);
1311 br_s.system_id = eth_addr_to_uint64(br->ea);
1312 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1313 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1316 br_s.system_id = eth_addr_to_uint64(br->ea);
1319 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1321 br_s.priority = strtoul(config_str, NULL, 0);
1323 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1326 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1328 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1330 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1333 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1335 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1337 br_s.max_age = STP_DEFAULT_MAX_AGE;
1340 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1342 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1344 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1347 /* Configure STP on the bridge. */
1348 if (ofproto_set_stp(br->ofproto, &br_s)) {
1349 VLOG_ERR("bridge %s: could not enable STP", br->name);
1353 /* Users must either set the port number with the "stp-port-num"
1354 * configuration on all ports or none. If manual configuration
1355 * is not done, then we allocate them sequentially. */
1356 port_num_counter = 0;
1357 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1358 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1359 struct ofproto_port_stp_settings port_s;
1360 struct iface *iface;
1362 port_configure_stp(br->ofproto, port, &port_s,
1363 &port_num_counter, port_num_bitmap);
1365 /* As bonds are not supported, just apply configuration to
1366 * all interfaces. */
1367 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1368 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1370 VLOG_ERR("port %s: could not enable STP", port->name);
1376 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1377 && port_num_counter) {
1378 VLOG_ERR("bridge %s: must manually configure all STP port "
1379 "IDs or none, disabling", br->name);
1380 ofproto_set_stp(br->ofproto, NULL);
1382 bitmap_free(port_num_bitmap);
1387 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1389 const struct port *port = port_lookup(br, name);
1390 return port && port_is_bond_fake_iface(port);
1394 port_is_bond_fake_iface(const struct port *port)
1396 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1400 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1402 struct bridge *br, *next;
1403 struct shash new_br;
1406 /* Collect new bridges' names and types. */
1407 shash_init(&new_br);
1408 for (i = 0; i < cfg->n_bridges; i++) {
1409 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1410 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1412 if (strchr(br_cfg->name, '/')) {
1413 /* Prevent remote ovsdb-server users from accessing arbitrary
1414 * directories, e.g. consider a bridge named "../../../etc/". */
1415 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1417 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1418 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1422 /* Get rid of deleted bridges or those whose types have changed.
1423 * Update 'cfg' of bridges that still exist. */
1424 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1425 br->cfg = shash_find_data(&new_br, br->name);
1426 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1427 br->cfg->datapath_type))) {
1432 /* Add new bridges. */
1433 for (i = 0; i < cfg->n_bridges; i++) {
1434 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1435 struct bridge *br = bridge_lookup(br_cfg->name);
1437 bridge_create(br_cfg);
1441 shash_destroy(&new_br);
1444 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1445 * Returns 0 if successful, otherwise a positive errno value. */
1447 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1448 struct netdev *netdev, char **errp)
1450 return netdev_set_config(netdev, &iface_cfg->options, errp);
1453 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1454 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1456 * If successful, returns 0 and stores the network device in '*netdevp'. On
1457 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1459 iface_do_create(const struct bridge *br,
1460 const struct ovsrec_interface *iface_cfg,
1461 const struct ovsrec_port *port_cfg,
1462 ofp_port_t *ofp_portp, struct netdev **netdevp,
1465 struct netdev *netdev = NULL;
1468 if (netdev_is_reserved_name(iface_cfg->name)) {
1469 VLOG_WARN("could not create interface %s, name is reserved",
1475 error = netdev_open(iface_cfg->name,
1476 iface_get_type(iface_cfg, br->cfg), &netdev);
1478 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1479 iface_cfg->name, ovs_strerror(error));
1483 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1488 *ofp_portp = iface_pick_ofport(iface_cfg);
1489 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1494 VLOG_INFO("bridge %s: added interface %s on port %d",
1495 br->name, iface_cfg->name, *ofp_portp);
1497 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1498 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1506 netdev_close(netdev);
1510 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1511 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1512 * automatically allocated for the iface. Takes ownership of and
1513 * deallocates 'if_cfg'.
1515 * Return true if an iface is successfully created, false otherwise. */
1517 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1518 const struct ovsrec_port *port_cfg)
1520 struct netdev *netdev;
1521 struct iface *iface;
1522 ofp_port_t ofp_port;
1527 /* Do the bits that can fail up front. */
1528 ovs_assert(!iface_lookup(br, iface_cfg->name));
1529 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1531 iface_clear_db_record(iface_cfg, errp);
1536 /* Get or create the port structure. */
1537 port = port_lookup(br, port_cfg->name);
1539 port = port_create(br, port_cfg);
1542 /* Create the iface structure. */
1543 iface = xzalloc(sizeof *iface);
1544 list_push_back(&port->ifaces, &iface->port_elem);
1545 hmap_insert(&br->iface_by_name, &iface->name_node,
1546 hash_string(iface_cfg->name, 0));
1548 iface->name = xstrdup(iface_cfg->name);
1549 iface->ofp_port = ofp_port;
1550 iface->netdev = netdev;
1551 iface->type = iface_get_type(iface_cfg, br->cfg);
1552 iface->cfg = iface_cfg;
1553 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1554 hash_ofp_port(ofp_port));
1556 /* Populate initial status in database. */
1557 iface_refresh_stats(iface);
1558 iface_refresh_netdev_status(iface);
1560 /* Add bond fake iface if necessary. */
1561 if (port_is_bond_fake_iface(port)) {
1562 struct ofproto_port ofproto_port;
1564 if (ofproto_port_query_by_name(br->ofproto, port->name,
1566 struct netdev *netdev;
1569 error = netdev_open(port->name, "internal", &netdev);
1571 ofp_port_t fake_ofp_port = OFPP_NONE;
1572 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1573 netdev_close(netdev);
1575 VLOG_WARN("could not open network device %s (%s)",
1576 port->name, ovs_strerror(error));
1579 /* Already exists, nothing to do. */
1580 ofproto_port_destroy(&ofproto_port);
1587 /* Set forward BPDU option. */
1589 bridge_configure_forward_bpdu(struct bridge *br)
1591 ofproto_set_forward_bpdu(br->ofproto,
1592 smap_get_bool(&br->cfg->other_config,
1597 /* Set MAC learning table configuration for 'br'. */
1599 bridge_configure_mac_table(struct bridge *br)
1601 const char *idle_time_str;
1604 const char *mac_table_size_str;
1607 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1608 idle_time = (idle_time_str && atoi(idle_time_str)
1609 ? atoi(idle_time_str)
1610 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1612 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1613 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1614 ? atoi(mac_table_size_str)
1617 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1620 /* Set multicast snooping table configuration for 'br'. */
1622 bridge_configure_mcast_snooping(struct bridge *br)
1624 if (!br->cfg->mcast_snooping_enable) {
1625 ofproto_set_mcast_snooping(br->ofproto, NULL);
1628 struct ofproto_mcast_snooping_settings br_s;
1629 const char *idle_time_str;
1630 const char *max_entries_str;
1632 idle_time_str = smap_get(&br->cfg->other_config,
1633 "mcast-snooping-aging-time");
1634 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1635 ? atoi(idle_time_str)
1636 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1638 max_entries_str = smap_get(&br->cfg->other_config,
1639 "mcast-snooping-table-size");
1640 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1641 ? atoi(max_entries_str)
1642 : MCAST_DEFAULT_MAX_ENTRIES);
1644 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1645 "mcast-snooping-disable-flood-unregistered",
1648 /* Configure multicast snooping on the bridge */
1649 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1650 VLOG_ERR("bridge %s: could not enable multicast snooping",
1655 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1656 bool flood = smap_get_bool(&port->cfg->other_config,
1657 "mcast-snooping-flood", false);
1658 if (ofproto_port_set_mcast_snooping(br->ofproto, port, flood)) {
1659 VLOG_ERR("port %s: could not configure mcast snooping",
1667 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1668 const struct port *fake_br, struct iface **hw_addr_iface)
1670 struct hmapx mirror_output_ports;
1672 bool found_addr = false;
1676 /* Mirror output ports don't participate in picking the local hardware
1677 * address. ofproto can't help us find out whether a given port is a
1678 * mirror output because we haven't configured mirrors yet, so we need to
1679 * accumulate them ourselves. */
1680 hmapx_init(&mirror_output_ports);
1681 for (i = 0; i < br->cfg->n_mirrors; i++) {
1682 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1683 if (m->output_port) {
1684 hmapx_add(&mirror_output_ports, m->output_port);
1688 /* Otherwise choose the minimum non-local MAC address among all of the
1690 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1691 uint8_t iface_ea[ETH_ADDR_LEN];
1692 struct iface *candidate;
1693 struct iface *iface;
1695 /* Mirror output ports don't participate. */
1696 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1700 /* Choose the MAC address to represent the port. */
1702 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1703 /* Find the interface with this Ethernet address (if any) so that
1704 * we can provide the correct devname to the caller. */
1705 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1706 uint8_t candidate_ea[ETH_ADDR_LEN];
1707 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1708 && eth_addr_equals(iface_ea, candidate_ea)) {
1713 /* Choose the interface whose MAC address will represent the port.
1714 * The Linux kernel bonding code always chooses the MAC address of
1715 * the first slave added to a bond, and the Fedora networking
1716 * scripts always add slaves to a bond in alphabetical order, so
1717 * for compatibility we choose the interface with the name that is
1718 * first in alphabetical order. */
1719 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1720 if (!iface || strcmp(candidate->name, iface->name) < 0) {
1725 /* The local port doesn't count (since we're trying to choose its
1726 * MAC address anyway). */
1727 if (iface->ofp_port == OFPP_LOCAL) {
1731 /* For fake bridges we only choose from ports with the same tag */
1732 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
1733 if (!port->cfg->tag) {
1736 if (*port->cfg->tag != *fake_br->cfg->tag) {
1742 error = netdev_get_etheraddr(iface->netdev, iface_ea);
1748 /* Compare against our current choice. */
1749 if (!eth_addr_is_multicast(iface_ea) &&
1750 !eth_addr_is_local(iface_ea) &&
1751 !eth_addr_is_reserved(iface_ea) &&
1752 !eth_addr_is_zero(iface_ea) &&
1753 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
1755 memcpy(ea, iface_ea, ETH_ADDR_LEN);
1756 *hw_addr_iface = iface;
1762 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
1763 *hw_addr_iface = NULL;
1766 hmapx_destroy(&mirror_output_ports);
1770 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1771 struct iface **hw_addr_iface)
1774 *hw_addr_iface = NULL;
1776 /* Did the user request a particular MAC? */
1777 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
1778 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
1779 if (eth_addr_is_multicast(ea)) {
1780 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1781 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
1782 } else if (eth_addr_is_zero(ea)) {
1783 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
1789 /* Find a local hw address */
1790 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
1793 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1794 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1795 * an interface on 'br', then that interface must be passed in as
1796 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1797 * 'hw_addr_iface' must be passed in as a null pointer. */
1799 bridge_pick_datapath_id(struct bridge *br,
1800 const uint8_t bridge_ea[ETH_ADDR_LEN],
1801 struct iface *hw_addr_iface)
1804 * The procedure for choosing a bridge MAC address will, in the most
1805 * ordinary case, also choose a unique MAC that we can use as a datapath
1806 * ID. In some special cases, though, multiple bridges will end up with
1807 * the same MAC address. This is OK for the bridges, but it will confuse
1808 * the OpenFlow controller, because each datapath needs a unique datapath
1811 * Datapath IDs must be unique. It is also very desirable that they be
1812 * stable from one run to the next, so that policy set on a datapath
1815 const char *datapath_id;
1818 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
1819 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
1823 if (!hw_addr_iface) {
1825 * A purely internal bridge, that is, one that has no non-virtual
1826 * network devices on it at all, is difficult because it has no
1827 * natural unique identifier at all.
1829 * When the host is a XenServer, we handle this case by hashing the
1830 * host's UUID with the name of the bridge. Names of bridges are
1831 * persistent across XenServer reboots, although they can be reused if
1832 * an internal network is destroyed and then a new one is later
1833 * created, so this is fairly effective.
1835 * When the host is not a XenServer, we punt by using a random MAC
1836 * address on each run.
1838 const char *host_uuid = xenserver_get_host_uuid();
1840 char *combined = xasprintf("%s,%s", host_uuid, br->name);
1841 dpid = dpid_from_hash(combined, strlen(combined));
1847 return eth_addr_to_uint64(bridge_ea);
1851 dpid_from_hash(const void *data, size_t n)
1853 uint8_t hash[SHA1_DIGEST_SIZE];
1855 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
1856 sha1_bytes(data, n, hash);
1857 eth_addr_mark_random(hash);
1858 return eth_addr_to_uint64(hash);
1862 iface_refresh_netdev_status(struct iface *iface)
1866 enum netdev_features current;
1867 enum netdev_flags flags;
1868 const char *link_state;
1869 uint8_t mac[ETH_ADDR_LEN];
1870 int64_t bps, mtu_64, ifindex64, link_resets;
1873 if (iface_is_synthetic(iface)) {
1877 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
1878 && !status_txn_try_again) {
1882 iface->change_seq = netdev_get_change_seq(iface->netdev);
1886 if (!netdev_get_status(iface->netdev, &smap)) {
1887 ovsrec_interface_set_status(iface->cfg, &smap);
1889 ovsrec_interface_set_status(iface->cfg, NULL);
1892 smap_destroy(&smap);
1894 error = netdev_get_flags(iface->netdev, &flags);
1896 const char *state = flags & NETDEV_UP ? "up" : "down";
1898 ovsrec_interface_set_admin_state(iface->cfg, state);
1900 ovsrec_interface_set_admin_state(iface->cfg, NULL);
1903 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
1904 ovsrec_interface_set_link_state(iface->cfg, link_state);
1906 link_resets = netdev_get_carrier_resets(iface->netdev);
1907 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
1909 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1910 bps = !error ? netdev_features_to_bps(current, 0) : 0;
1912 ovsrec_interface_set_duplex(iface->cfg,
1913 netdev_features_is_full_duplex(current)
1915 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
1917 ovsrec_interface_set_duplex(iface->cfg, NULL);
1918 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
1921 error = netdev_get_mtu(iface->netdev, &mtu);
1924 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
1926 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
1929 error = netdev_get_etheraddr(iface->netdev, mac);
1931 char mac_string[32];
1933 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
1934 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
1936 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
1939 /* The netdev may return a negative number (such as -EOPNOTSUPP)
1940 * if there is no valid ifindex number. */
1941 ifindex64 = netdev_get_ifindex(iface->netdev);
1942 if (ifindex64 < 0) {
1945 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
1949 iface_refresh_ofproto_status(struct iface *iface)
1953 if (iface_is_synthetic(iface)) {
1957 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
1961 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
1963 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
1966 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
1968 || status_txn_try_again) {
1969 iface_refresh_cfm_stats(iface);
1972 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
1974 || status_txn_try_again) {
1978 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
1979 iface->ofp_port, &smap);
1980 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
1981 smap_destroy(&smap);
1985 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
1988 iface_refresh_cfm_stats(struct iface *iface)
1990 const struct ovsrec_interface *cfg = iface->cfg;
1991 struct cfm_status status;
1994 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
1995 iface->ofp_port, &status);
1997 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
1998 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
1999 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2000 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2001 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2002 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2004 const char *reasons[CFM_FAULT_N_REASONS];
2005 int64_t cfm_health = status.health;
2006 int64_t cfm_flap_count = status.flap_count;
2007 bool faulted = status.faults != 0;
2010 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2013 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2014 int reason = 1 << i;
2015 if (status.faults & reason) {
2016 reasons[j++] = cfm_fault_reason_to_str(reason);
2019 ovsrec_interface_set_cfm_fault_status(cfg, (char **) reasons, j);
2021 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2023 if (status.remote_opstate >= 0) {
2024 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2025 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2027 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2030 ovsrec_interface_set_cfm_remote_mpids(cfg,
2031 (const int64_t *)status.rmps,
2033 if (cfm_health >= 0) {
2034 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2036 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2044 iface_refresh_stats(struct iface *iface)
2046 #define IFACE_STATS \
2047 IFACE_STAT(rx_packets, "rx_packets") \
2048 IFACE_STAT(tx_packets, "tx_packets") \
2049 IFACE_STAT(rx_bytes, "rx_bytes") \
2050 IFACE_STAT(tx_bytes, "tx_bytes") \
2051 IFACE_STAT(rx_dropped, "rx_dropped") \
2052 IFACE_STAT(tx_dropped, "tx_dropped") \
2053 IFACE_STAT(rx_errors, "rx_errors") \
2054 IFACE_STAT(tx_errors, "tx_errors") \
2055 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2056 IFACE_STAT(rx_over_errors, "rx_over_err") \
2057 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2058 IFACE_STAT(collisions, "collisions")
2060 #define IFACE_STAT(MEMBER, NAME) + 1
2061 enum { N_IFACE_STATS = IFACE_STATS };
2063 int64_t values[N_IFACE_STATS];
2064 char *keys[N_IFACE_STATS];
2067 struct netdev_stats stats;
2069 if (iface_is_synthetic(iface)) {
2073 /* Intentionally ignore return value, since errors will set 'stats' to
2074 * all-1s, and we will deal with that correctly below. */
2075 netdev_get_stats(iface->netdev, &stats);
2077 /* Copy statistics into keys[] and values[]. */
2079 #define IFACE_STAT(MEMBER, NAME) \
2080 if (stats.MEMBER != UINT64_MAX) { \
2082 values[n] = stats.MEMBER; \
2087 ovs_assert(n <= N_IFACE_STATS);
2089 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2094 br_refresh_stp_status(struct bridge *br)
2096 struct smap smap = SMAP_INITIALIZER(&smap);
2097 struct ofproto *ofproto = br->ofproto;
2098 struct ofproto_stp_status status;
2100 if (ofproto_get_stp_status(ofproto, &status)) {
2104 if (!status.enabled) {
2105 ovsrec_bridge_set_status(br->cfg, NULL);
2109 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2110 STP_ID_ARGS(status.bridge_id));
2111 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2112 STP_ID_ARGS(status.designated_root));
2113 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2115 ovsrec_bridge_set_status(br->cfg, &smap);
2116 smap_destroy(&smap);
2120 port_refresh_stp_status(struct port *port)
2122 struct ofproto *ofproto = port->bridge->ofproto;
2123 struct iface *iface;
2124 struct ofproto_port_stp_status status;
2127 if (port_is_synthetic(port)) {
2131 /* STP doesn't currently support bonds. */
2132 if (!list_is_singleton(&port->ifaces)) {
2133 ovsrec_port_set_status(port->cfg, NULL);
2137 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2138 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2142 if (!status.enabled) {
2143 ovsrec_port_set_status(port->cfg, NULL);
2147 /* Set Status column. */
2149 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2150 smap_add(&smap, "stp_state", stp_state_name(status.state));
2151 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2152 smap_add(&smap, "stp_role", stp_role_name(status.role));
2153 ovsrec_port_set_status(port->cfg, &smap);
2154 smap_destroy(&smap);
2158 port_refresh_stp_stats(struct port *port)
2160 struct ofproto *ofproto = port->bridge->ofproto;
2161 struct iface *iface;
2162 struct ofproto_port_stp_stats stats;
2164 int64_t int_values[3];
2166 if (port_is_synthetic(port)) {
2170 /* STP doesn't currently support bonds. */
2171 if (!list_is_singleton(&port->ifaces)) {
2175 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2176 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2180 if (!stats.enabled) {
2181 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2185 /* Set Statistics column. */
2186 keys[0] = "stp_tx_count";
2187 int_values[0] = stats.tx_count;
2188 keys[1] = "stp_rx_count";
2189 int_values[1] = stats.rx_count;
2190 keys[2] = "stp_error_count";
2191 int_values[2] = stats.error_count;
2193 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2194 ARRAY_SIZE(int_values));
2198 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2200 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2204 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2206 bool enable = enable_system_stats(cfg);
2208 system_stats_enable(enable);
2210 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2215 run_system_stats(void)
2217 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2220 stats = system_stats_run();
2222 struct ovsdb_idl_txn *txn;
2223 struct ovsdb_datum datum;
2225 txn = ovsdb_idl_txn_create(idl);
2226 ovsdb_datum_from_smap(&datum, stats);
2227 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2229 ovsdb_idl_txn_commit(txn);
2230 ovsdb_idl_txn_destroy(txn);
2237 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2240 case OFPCR12_ROLE_EQUAL:
2242 case OFPCR12_ROLE_MASTER:
2244 case OFPCR12_ROLE_SLAVE:
2246 case OFPCR12_ROLE_NOCHANGE:
2248 return "*** INVALID ROLE ***";
2253 refresh_controller_status(void)
2257 const struct ovsrec_controller *cfg;
2261 /* Accumulate status for controllers on all bridges. */
2262 HMAP_FOR_EACH (br, node, &all_bridges) {
2263 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2266 /* Update each controller in the database with current status. */
2267 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2268 struct ofproto_controller_info *cinfo =
2269 shash_find_data(&info, cfg->target);
2272 struct smap smap = SMAP_INITIALIZER(&smap);
2273 const char **values = cinfo->pairs.values;
2274 const char **keys = cinfo->pairs.keys;
2277 for (i = 0; i < cinfo->pairs.n; i++) {
2278 smap_add(&smap, keys[i], values[i]);
2281 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2282 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2284 ovsrec_controller_set_status(cfg, &smap);
2285 smap_destroy(&smap);
2287 ovsrec_controller_set_is_connected(cfg, false);
2288 ovsrec_controller_set_role(cfg, NULL);
2289 ovsrec_controller_set_status(cfg, NULL);
2293 ofproto_free_ofproto_controller_info(&info);
2303 /* Let each datapath type do the work that it needs to do. */
2305 ofproto_enumerate_types(&types);
2306 SSET_FOR_EACH (type, &types) {
2307 ofproto_type_run(type);
2309 sset_destroy(&types);
2311 /* Let each bridge do the work that it needs to do. */
2312 HMAP_FOR_EACH (br, node, &all_bridges) {
2313 ofproto_run(br->ofproto);
2320 static struct ovsrec_open_vswitch null_cfg;
2321 const struct ovsrec_open_vswitch *cfg;
2323 bool vlan_splinters_changed;
2327 ovsrec_open_vswitch_init(&null_cfg);
2331 if (ovsdb_idl_is_lock_contended(idl)) {
2332 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2333 struct bridge *br, *next_br;
2335 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2336 "disabling this process (pid %ld) until it goes away",
2337 (long int) getpid());
2339 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2342 /* Since we will not be running system_stats_run() in this process
2343 * with the current situation of multiple ovs-vswitchd daemons,
2344 * disable system stats collection. */
2345 system_stats_enable(false);
2347 } else if (!ovsdb_idl_has_lock(idl)) {
2350 cfg = ovsrec_open_vswitch_first(idl);
2352 /* Initialize the ofproto library. This only needs to run once, but
2353 * it must be done after the configuration is set. If the
2354 * initialization has already occurred, bridge_init_ofproto()
2355 * returns immediately. */
2356 bridge_init_ofproto(cfg);
2358 /* Once the value of flow-restore-wait is false, we no longer should
2359 * check its value from the database. */
2360 if (cfg && ofproto_get_flow_restore_wait()) {
2361 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2362 "flow-restore-wait", false));
2367 /* Re-configure SSL. We do this on every trip through the main loop,
2368 * instead of just when the database changes, because the contents of the
2369 * key and certificate files can change without the database changing.
2371 * We do this before bridge_reconfigure() because that function might
2372 * initiate SSL connections and thus requires SSL to be configured. */
2373 if (cfg && cfg->ssl) {
2374 const struct ovsrec_ssl *ssl = cfg->ssl;
2376 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2377 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2380 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2381 * usage has changed. */
2382 vlan_splinters_changed = false;
2383 if (vlan_splinters_enabled_anywhere) {
2384 HMAP_FOR_EACH (br, node, &all_bridges) {
2385 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2386 vlan_splinters_changed = true;
2392 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2393 struct ovsdb_idl_txn *txn;
2395 idl_seqno = ovsdb_idl_get_seqno(idl);
2396 txn = ovsdb_idl_txn_create(idl);
2397 bridge_reconfigure(cfg ? cfg : &null_cfg);
2400 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2403 /* If we are completing our initial configuration for this run
2404 * of ovs-vswitchd, then keep the transaction around to monitor
2405 * it for completion. */
2406 if (initial_config_done) {
2407 /* Always sets the 'status_txn_try_again' to check again,
2408 * in case that this transaction fails. */
2409 status_txn_try_again = true;
2410 ovsdb_idl_txn_commit(txn);
2411 ovsdb_idl_txn_destroy(txn);
2413 initial_config_done = true;
2414 daemonize_txn = txn;
2418 if (daemonize_txn) {
2419 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2420 if (status != TXN_INCOMPLETE) {
2421 ovsdb_idl_txn_destroy(daemonize_txn);
2422 daemonize_txn = NULL;
2424 /* ovs-vswitchd has completed initialization, so allow the
2425 * process that forked us to exit successfully. */
2426 daemonize_complete();
2428 vlog_enable_async();
2430 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
2434 /* Statistics update interval should always be greater than or equal to
2437 stats_interval = MAX(smap_get_int(&cfg->other_config,
2438 "stats-update-interval",
2441 stats_interval = 5000;
2443 if (stats_timer_interval != stats_interval) {
2444 stats_timer_interval = stats_interval;
2445 stats_timer = LLONG_MIN;
2448 /* Refresh interface and mirror stats if necessary. */
2449 if (time_msec() >= stats_timer) {
2451 struct ovsdb_idl_txn *txn;
2453 txn = ovsdb_idl_txn_create(idl);
2454 HMAP_FOR_EACH (br, node, &all_bridges) {
2458 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2459 struct iface *iface;
2461 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2462 iface_refresh_stats(iface);
2465 port_refresh_stp_stats(port);
2468 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2469 mirror_refresh_stats(m);
2473 refresh_controller_status();
2474 ovsdb_idl_txn_commit(txn);
2475 ovsdb_idl_txn_destroy(txn); /* XXX */
2478 stats_timer = time_msec() + stats_timer_interval;
2484 /* Check the need to update status. */
2485 seq = seq_read(connectivity_seq_get());
2486 if (seq != connectivity_seqno || status_txn_try_again) {
2487 connectivity_seqno = seq;
2488 status_txn = ovsdb_idl_txn_create(idl);
2489 HMAP_FOR_EACH (br, node, &all_bridges) {
2492 br_refresh_stp_status(br);
2493 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2494 struct iface *iface;
2496 port_refresh_stp_status(port);
2497 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2498 iface_refresh_netdev_status(iface);
2499 iface_refresh_ofproto_status(iface);
2507 enum ovsdb_idl_txn_status status;
2509 status = ovsdb_idl_txn_commit(status_txn);
2510 /* Do not destroy "status_txn" if the transaction is
2511 * "TXN_INCOMPLETE". */
2512 if (status != TXN_INCOMPLETE) {
2513 ovsdb_idl_txn_destroy(status_txn);
2516 /* Sets the 'status_txn_try_again' if the transaction fails. */
2517 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2518 status_txn_try_again = false;
2520 status_txn_try_again = true;
2534 ovsdb_idl_wait(idl);
2535 if (daemonize_txn) {
2536 ovsdb_idl_txn_wait(daemonize_txn);
2540 ofproto_enumerate_types(&types);
2541 SSET_FOR_EACH (type, &types) {
2542 ofproto_type_wait(type);
2544 sset_destroy(&types);
2546 if (!hmap_is_empty(&all_bridges)) {
2549 HMAP_FOR_EACH (br, node, &all_bridges) {
2550 ofproto_wait(br->ofproto);
2553 poll_timer_wait_until(stats_timer);
2556 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2557 * transaction to complete. If the status update to database needs to be
2558 * run again (transaction fails), registers a timeout in
2559 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2560 * sequence number. */
2562 ovsdb_idl_txn_wait(status_txn);
2563 } else if (status_txn_try_again) {
2564 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2566 seq_wait(connectivity_seq_get(), connectivity_seqno);
2569 system_stats_wait();
2572 /* Adds some memory usage statistics for bridges into 'usage', for use with
2573 * memory_report(). */
2575 bridge_get_memory_usage(struct simap *usage)
2582 ofproto_enumerate_types(&types);
2583 SSET_FOR_EACH (type, &types) {
2584 ofproto_type_get_memory_usage(type, usage);
2586 sset_destroy(&types);
2588 HMAP_FOR_EACH (br, node, &all_bridges) {
2589 ofproto_get_memory_usage(br->ofproto, usage);
2593 /* QoS unixctl user interface functions. */
2595 struct qos_unixctl_show_cbdata {
2597 struct iface *iface;
2601 qos_unixctl_show_queue(unsigned int queue_id,
2602 const struct smap *details,
2603 struct iface *iface,
2606 struct netdev_queue_stats stats;
2607 struct smap_node *node;
2610 ds_put_cstr(ds, "\n");
2612 ds_put_format(ds, "Queue %u:\n", queue_id);
2614 ds_put_cstr(ds, "Default:\n");
2617 SMAP_FOR_EACH (node, details) {
2618 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
2621 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
2623 if (stats.tx_packets != UINT64_MAX) {
2624 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
2627 if (stats.tx_bytes != UINT64_MAX) {
2628 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
2631 if (stats.tx_errors != UINT64_MAX) {
2632 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
2635 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
2636 queue_id, ovs_strerror(error));
2641 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
2642 const char *argv[], void *aux OVS_UNUSED)
2644 struct ds ds = DS_EMPTY_INITIALIZER;
2645 struct smap smap = SMAP_INITIALIZER(&smap);
2646 struct iface *iface;
2648 struct smap_node *node;
2650 iface = iface_find(argv[1]);
2652 unixctl_command_reply_error(conn, "no such interface");
2656 netdev_get_qos(iface->netdev, &type, &smap);
2658 if (*type != '\0') {
2659 struct netdev_queue_dump dump;
2660 struct smap details;
2661 unsigned int queue_id;
2663 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
2665 SMAP_FOR_EACH (node, &smap) {
2666 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
2669 smap_init(&details);
2670 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
2671 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
2673 smap_destroy(&details);
2675 unixctl_command_reply(conn, ds_cstr(&ds));
2677 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
2678 unixctl_command_reply_error(conn, ds_cstr(&ds));
2681 smap_destroy(&smap);
2685 /* Bridge reconfiguration functions. */
2687 bridge_create(const struct ovsrec_bridge *br_cfg)
2691 ovs_assert(!bridge_lookup(br_cfg->name));
2692 br = xzalloc(sizeof *br);
2694 br->name = xstrdup(br_cfg->name);
2695 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
2698 /* Derive the default Ethernet address from the bridge's UUID. This should
2699 * be unique and it will be stable between ovs-vswitchd runs. */
2700 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
2701 eth_addr_mark_random(br->default_ea);
2703 hmap_init(&br->ports);
2704 hmap_init(&br->ifaces);
2705 hmap_init(&br->iface_by_name);
2706 hmap_init(&br->mirrors);
2708 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
2712 bridge_destroy(struct bridge *br)
2715 struct mirror *mirror, *next_mirror;
2716 struct port *port, *next_port;
2718 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
2721 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
2722 mirror_destroy(mirror);
2725 hmap_remove(&all_bridges, &br->node);
2726 ofproto_destroy(br->ofproto);
2727 hmap_destroy(&br->ifaces);
2728 hmap_destroy(&br->ports);
2729 hmap_destroy(&br->iface_by_name);
2730 hmap_destroy(&br->mirrors);
2737 static struct bridge *
2738 bridge_lookup(const char *name)
2742 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
2743 if (!strcmp(br->name, name)) {
2750 /* Handle requests for a listing of all flows known by the OpenFlow
2751 * stack, including those normally hidden. */
2753 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
2754 const char *argv[], void *aux OVS_UNUSED)
2759 br = bridge_lookup(argv[1]);
2761 unixctl_command_reply_error(conn, "Unknown bridge");
2766 ofproto_get_all_flows(br->ofproto, &results);
2768 unixctl_command_reply(conn, ds_cstr(&results));
2769 ds_destroy(&results);
2772 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
2773 * connections and reconnect. If BRIDGE is not specified, then all bridges
2774 * drop their controller connections and reconnect. */
2776 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
2777 const char *argv[], void *aux OVS_UNUSED)
2781 br = bridge_lookup(argv[1]);
2783 unixctl_command_reply_error(conn, "Unknown bridge");
2786 ofproto_reconnect_controllers(br->ofproto);
2788 HMAP_FOR_EACH (br, node, &all_bridges) {
2789 ofproto_reconnect_controllers(br->ofproto);
2792 unixctl_command_reply(conn, NULL);
2796 bridge_get_controllers(const struct bridge *br,
2797 struct ovsrec_controller ***controllersp)
2799 struct ovsrec_controller **controllers;
2800 size_t n_controllers;
2802 controllers = br->cfg->controller;
2803 n_controllers = br->cfg->n_controller;
2805 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
2811 *controllersp = controllers;
2813 return n_controllers;
2817 bridge_collect_wanted_ports(struct bridge *br,
2818 const unsigned long int *splinter_vlans,
2819 struct shash *wanted_ports)
2823 shash_init(wanted_ports);
2825 for (i = 0; i < br->cfg->n_ports; i++) {
2826 const char *name = br->cfg->ports[i]->name;
2827 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
2828 VLOG_WARN("bridge %s: %s specified twice as bridge port",
2832 if (bridge_get_controllers(br, NULL)
2833 && !shash_find(wanted_ports, br->name)) {
2834 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
2835 br->name, br->name);
2837 ovsrec_interface_init(&br->synth_local_iface);
2838 ovsrec_port_init(&br->synth_local_port);
2840 br->synth_local_port.interfaces = &br->synth_local_ifacep;
2841 br->synth_local_port.n_interfaces = 1;
2842 br->synth_local_port.name = br->name;
2844 br->synth_local_iface.name = br->name;
2845 br->synth_local_iface.type = "internal";
2847 br->synth_local_ifacep = &br->synth_local_iface;
2849 shash_add(wanted_ports, br->name, &br->synth_local_port);
2852 if (splinter_vlans) {
2853 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
2857 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
2858 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
2859 * 'br' needs to complete its configuration. */
2861 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
2863 struct shash_node *port_node;
2864 struct port *port, *next;
2866 /* Get rid of deleted ports.
2867 * Get rid of deleted interfaces on ports that still exist. */
2868 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
2869 port->cfg = shash_find_data(wanted_ports, port->name);
2873 port_del_ifaces(port);
2877 /* Update iface->cfg and iface->type in interfaces that still exist. */
2878 SHASH_FOR_EACH (port_node, wanted_ports) {
2879 const struct ovsrec_port *port = port_node->data;
2882 for (i = 0; i < port->n_interfaces; i++) {
2883 const struct ovsrec_interface *cfg = port->interfaces[i];
2884 struct iface *iface = iface_lookup(br, cfg->name);
2885 const char *type = iface_get_type(cfg, br->cfg);
2890 } else if (!strcmp(type, "null")) {
2891 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
2892 " may be removed in February 2013. Please email"
2893 " dev@openvswitch.org with concerns.",
2896 /* We will add new interfaces later. */
2902 /* Initializes 'oc' appropriately as a management service controller for
2905 * The caller must free oc->target when it is no longer needed. */
2907 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
2908 struct ofproto_controller *oc)
2910 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
2911 oc->max_backoff = 0;
2912 oc->probe_interval = 60;
2913 oc->band = OFPROTO_OUT_OF_BAND;
2915 oc->burst_limit = 0;
2916 oc->enable_async_msgs = true;
2920 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
2922 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
2923 struct ofproto_controller *oc)
2927 oc->target = c->target;
2928 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
2929 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
2930 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
2931 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
2932 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
2933 oc->burst_limit = (c->controller_burst_limit
2934 ? *c->controller_burst_limit : 0);
2935 oc->enable_async_msgs = (!c->enable_async_messages
2936 || *c->enable_async_messages);
2937 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
2938 if (dscp < 0 || dscp > 63) {
2939 dscp = DSCP_DEFAULT;
2944 /* Configures the IP stack for 'br''s local interface properly according to the
2945 * configuration in 'c'. */
2947 bridge_configure_local_iface_netdev(struct bridge *br,
2948 struct ovsrec_controller *c)
2950 struct netdev *netdev;
2951 struct in_addr mask, gateway;
2953 struct iface *local_iface;
2956 /* If there's no local interface or no IP address, give up. */
2957 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
2958 if (!local_iface || !c->local_ip
2959 || !inet_pton(AF_INET, c->local_ip, &ip)) {
2963 /* Bring up the local interface. */
2964 netdev = local_iface->netdev;
2965 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
2967 /* Configure the IP address and netmask. */
2968 if (!c->local_netmask
2969 || !inet_pton(AF_INET, c->local_netmask, &mask)
2971 mask.s_addr = guess_netmask(ip.s_addr);
2973 if (!netdev_set_in4(netdev, ip, mask)) {
2974 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
2975 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
2978 /* Configure the default gateway. */
2979 if (c->local_gateway
2980 && inet_pton(AF_INET, c->local_gateway, &gateway)
2981 && gateway.s_addr) {
2982 if (!netdev_add_router(netdev, gateway)) {
2983 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
2984 br->name, IP_ARGS(gateway.s_addr));
2989 /* Returns true if 'a' and 'b' are the same except that any number of slashes
2990 * in either string are treated as equal to any number of slashes in the other,
2991 * e.g. "x///y" is equal to "x/y".
2993 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
2994 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
2995 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
2996 * 'b' against a prefix of 'a'.
2999 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3001 const char *b_start = b;
3003 if (b - b_start >= b_stoplen) {
3005 } else if (*a != *b) {
3007 } else if (*a == '/') {
3008 a += strspn(a, "/");
3009 b += strspn(b, "/");
3010 } else if (*a == '\0') {
3020 bridge_configure_remotes(struct bridge *br,
3021 const struct sockaddr_in *managers, size_t n_managers)
3023 bool disable_in_band;
3025 struct ovsrec_controller **controllers;
3026 size_t n_controllers;
3028 enum ofproto_fail_mode fail_mode;
3030 struct ofproto_controller *ocs;
3034 /* Check if we should disable in-band control on this bridge. */
3035 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3038 /* Set OpenFlow queue ID for in-band control. */
3039 ofproto_set_in_band_queue(br->ofproto,
3040 smap_get_int(&br->cfg->other_config,
3041 "in-band-queue", -1));
3043 if (disable_in_band) {
3044 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3046 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3049 n_controllers = bridge_get_controllers(br, &controllers);
3051 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3054 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3055 for (i = 0; i < n_controllers; i++) {
3056 struct ovsrec_controller *c = controllers[i];
3058 if (!strncmp(c->target, "punix:", 6)
3059 || !strncmp(c->target, "unix:", 5)) {
3060 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3063 if (!strncmp(c->target, "unix:", 5)) {
3064 /* Connect to a listening socket */
3065 whitelist = xasprintf("unix:%s/", ovs_rundir());
3066 if (strchr(c->target, '/') &&
3067 !equal_pathnames(c->target, whitelist,
3068 strlen(whitelist))) {
3069 /* Absolute path specified, but not in ovs_rundir */
3070 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3071 "controller \"%s\" due to possibility for "
3072 "remote exploit. Instead, specify socket "
3073 "in whitelisted \"%s\" or connect to "
3074 "\"unix:%s/%s.mgmt\" (which is always "
3075 "available without special configuration).",
3076 br->name, c->target, whitelist,
3077 ovs_rundir(), br->name);
3082 whitelist = xasprintf("punix:%s/%s.controller",
3083 ovs_rundir(), br->name);
3084 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3085 /* Prevent remote ovsdb-server users from accessing
3086 * arbitrary Unix domain sockets and overwriting arbitrary
3088 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3089 "controller \"%s\" due to possibility of "
3090 "overwriting local files. Instead, specify "
3091 "whitelisted \"%s\" or connect to "
3092 "\"unix:%s/%s.mgmt\" (which is always "
3093 "available without special configuration).",
3094 br->name, c->target, whitelist,
3095 ovs_rundir(), br->name);
3104 bridge_configure_local_iface_netdev(br, c);
3105 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3106 if (disable_in_band) {
3107 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3112 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3113 bridge_get_allowed_versions(br));
3114 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3117 /* Set the fail-mode. */
3118 fail_mode = !br->cfg->fail_mode
3119 || !strcmp(br->cfg->fail_mode, "standalone")
3120 ? OFPROTO_FAIL_STANDALONE
3121 : OFPROTO_FAIL_SECURE;
3122 ofproto_set_fail_mode(br->ofproto, fail_mode);
3124 /* Configure OpenFlow controller connection snooping. */
3125 if (!ofproto_has_snoops(br->ofproto)) {
3129 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3130 ovs_rundir(), br->name));
3131 ofproto_set_snoops(br->ofproto, &snoops);
3132 sset_destroy(&snoops);
3137 bridge_configure_tables(struct bridge *br)
3139 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3143 n_tables = ofproto_get_n_tables(br->ofproto);
3145 for (i = 0; i < n_tables; i++) {
3146 struct ofproto_table_settings s;
3149 s.max_flows = UINT_MAX;
3152 s.n_prefix_fields = 0;
3153 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3155 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3156 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3160 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3161 s.max_flows = *cfg->flow_limit;
3163 if (cfg->overflow_policy
3164 && !strcmp(cfg->overflow_policy, "evict")) {
3166 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3167 for (k = 0; k < cfg->n_groups; k++) {
3168 const char *string = cfg->groups[k];
3171 msg = mf_parse_subfield__(&s.groups[k], &string);
3173 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3174 "'groups' (%s)", br->name, i, msg);
3176 } else if (*string) {
3177 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3178 "element '%s' contains trailing garbage",
3179 br->name, i, cfg->groups[k]);
3185 /* Prefix lookup fields. */
3186 no_prefixes = false;
3187 s.n_prefix_fields = 0;
3188 for (k = 0; k < cfg->n_prefixes; k++) {
3189 const char *name = cfg->prefixes[k];
3190 const struct mf_field *mf;
3192 if (strcmp(name, "none") == 0) {
3196 mf = mf_from_name(name);
3198 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3202 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3203 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3204 "%s", br->name, name);
3207 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3208 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3209 "field not used: %s", br->name, name);
3212 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3214 if (s.n_prefix_fields == 0 && !no_prefixes) {
3215 /* Use default values. */
3216 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3217 memcpy(s.prefix_fields, default_prefix_fields,
3218 sizeof default_prefix_fields);
3220 if (s.n_prefix_fields > 0) {
3222 struct ds ds = DS_EMPTY_INITIALIZER;
3223 for (k = 0; k < s.n_prefix_fields; k++) {
3225 ds_put_char(&ds, ',');
3227 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3229 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3230 br->name, i, ds_cstr(&ds));
3235 ofproto_configure_table(br->ofproto, i, &s);
3239 for (; j < br->cfg->n_flow_tables; j++) {
3240 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3241 "%"PRId64" not supported by this datapath", br->name,
3242 br->cfg->key_flow_tables[j]);
3247 bridge_configure_dp_desc(struct bridge *br)
3249 ofproto_set_dp_desc(br->ofproto,
3250 smap_get(&br->cfg->other_config, "dp-desc"));
3253 /* Port functions. */
3255 static struct port *
3256 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3260 port = xzalloc(sizeof *port);
3262 port->name = xstrdup(cfg->name);
3264 list_init(&port->ifaces);
3266 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3270 /* Deletes interfaces from 'port' that are no longer configured for it. */
3272 port_del_ifaces(struct port *port)
3274 struct iface *iface, *next;
3275 struct sset new_ifaces;
3278 /* Collect list of new interfaces. */
3279 sset_init(&new_ifaces);
3280 for (i = 0; i < port->cfg->n_interfaces; i++) {
3281 const char *name = port->cfg->interfaces[i]->name;
3282 const char *type = port->cfg->interfaces[i]->type;
3283 if (strcmp(type, "null")) {
3284 sset_add(&new_ifaces, name);
3288 /* Get rid of deleted interfaces. */
3289 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3290 if (!sset_contains(&new_ifaces, iface->name)) {
3291 iface_destroy(iface);
3295 sset_destroy(&new_ifaces);
3299 port_destroy(struct port *port)
3302 struct bridge *br = port->bridge;
3303 struct iface *iface, *next;
3306 ofproto_bundle_unregister(br->ofproto, port);
3309 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3310 iface_destroy__(iface);
3313 hmap_remove(&br->ports, &port->hmap_node);
3319 static struct port *
3320 port_lookup(const struct bridge *br, const char *name)
3324 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
3326 if (!strcmp(port->name, name)) {
3334 enable_lacp(struct port *port, bool *activep)
3336 if (!port->cfg->lacp) {
3337 /* XXX when LACP implementation has been sufficiently tested, enable by
3338 * default and make active on bonded ports. */
3340 } else if (!strcmp(port->cfg->lacp, "off")) {
3342 } else if (!strcmp(port->cfg->lacp, "active")) {
3345 } else if (!strcmp(port->cfg->lacp, "passive")) {
3349 VLOG_WARN("port %s: unknown LACP mode %s",
3350 port->name, port->cfg->lacp);
3355 static struct lacp_settings *
3356 port_configure_lacp(struct port *port, struct lacp_settings *s)
3358 const char *lacp_time, *system_id;
3361 if (!enable_lacp(port, &s->active)) {
3365 s->name = port->name;
3367 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
3369 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
3370 ETH_ADDR_SCAN_ARGS(s->id))) {
3371 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
3372 " address.", port->name, system_id);
3376 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
3379 if (eth_addr_is_zero(s->id)) {
3380 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
3384 /* Prefer bondable links if unspecified. */
3385 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
3387 s->priority = (priority > 0 && priority <= UINT16_MAX
3389 : UINT16_MAX - !list_is_short(&port->ifaces));
3391 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
3392 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
3394 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3395 "lacp-fallback-ab", false);
3401 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
3403 int priority, portid, key;
3405 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
3406 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
3408 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
3410 if (portid <= 0 || portid > UINT16_MAX) {
3411 portid = ofp_to_u16(iface->ofp_port);
3414 if (priority <= 0 || priority > UINT16_MAX) {
3415 priority = UINT16_MAX;
3418 if (key < 0 || key > UINT16_MAX) {
3422 s->name = iface->name;
3424 s->priority = priority;
3429 port_configure_bond(struct port *port, struct bond_settings *s)
3431 const char *detect_s;
3432 struct iface *iface;
3433 int miimon_interval;
3435 s->name = port->name;
3437 if (port->cfg->bond_mode) {
3438 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
3439 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
3440 port->name, port->cfg->bond_mode,
3441 bond_mode_to_string(s->balance));
3444 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
3446 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
3447 * active-backup. At some point we should remove this warning. */
3448 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
3449 " in previous versions, the default bond_mode was"
3450 " balance-slb", port->name,
3451 bond_mode_to_string(s->balance));
3453 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
3454 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
3455 "please use another bond type or disable flood_vlans",
3459 miimon_interval = smap_get_int(&port->cfg->other_config,
3460 "bond-miimon-interval", 0);
3461 if (miimon_interval <= 0) {
3462 miimon_interval = 200;
3465 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
3466 if (!detect_s || !strcmp(detect_s, "carrier")) {
3467 miimon_interval = 0;
3468 } else if (strcmp(detect_s, "miimon")) {
3469 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
3470 "defaulting to carrier", port->name, detect_s);
3471 miimon_interval = 0;
3474 s->up_delay = MAX(0, port->cfg->bond_updelay);
3475 s->down_delay = MAX(0, port->cfg->bond_downdelay);
3476 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
3477 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
3478 "bond-rebalance-interval", 10000);
3479 if (s->rebalance_interval && s->rebalance_interval < 1000) {
3480 s->rebalance_interval = 1000;
3483 s->fake_iface = port->cfg->bond_fake_iface;
3485 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
3486 "lacp-fallback-ab", false);
3488 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
3489 netdev_set_miimon_interval(iface->netdev, miimon_interval);
3493 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
3494 * instead of obtaining it from the database. */
3496 port_is_synthetic(const struct port *port)
3498 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
3501 /* Interface functions. */
3504 iface_is_internal(const struct ovsrec_interface *iface,
3505 const struct ovsrec_bridge *br)
3507 /* The local port and "internal" ports are always "internal". */
3508 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
3511 /* Returns the correct network device type for interface 'iface' in bridge
3514 iface_get_type(const struct ovsrec_interface *iface,
3515 const struct ovsrec_bridge *br)
3519 /* The local port always has type "internal". Other ports take
3520 * their type from the database and default to "system" if none is
3522 if (iface_is_internal(iface, br)) {
3525 type = iface->type[0] ? iface->type : "system";
3528 return ofproto_port_open_type(br->datapath_type, type);
3532 iface_destroy__(struct iface *iface)
3535 struct port *port = iface->port;
3536 struct bridge *br = port->bridge;
3538 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
3539 ofproto_port_unregister(br->ofproto, iface->ofp_port);
3542 if (iface->ofp_port != OFPP_NONE) {
3543 hmap_remove(&br->ifaces, &iface->ofp_port_node);
3546 list_remove(&iface->port_elem);
3547 hmap_remove(&br->iface_by_name, &iface->name_node);
3549 /* The user is changing configuration here, so netdev_remove needs to be
3550 * used as opposed to netdev_close */
3551 netdev_remove(iface->netdev);
3559 iface_destroy(struct iface *iface)
3562 struct port *port = iface->port;
3564 iface_destroy__(iface);
3565 if (list_is_empty(&port->ifaces)) {
3571 static struct iface *
3572 iface_lookup(const struct bridge *br, const char *name)
3574 struct iface *iface;
3576 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
3577 &br->iface_by_name) {
3578 if (!strcmp(iface->name, name)) {
3586 static struct iface *
3587 iface_find(const char *name)
3589 const struct bridge *br;
3591 HMAP_FOR_EACH (br, node, &all_bridges) {
3592 struct iface *iface = iface_lookup(br, name);
3601 static struct iface *
3602 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
3604 struct iface *iface;
3606 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
3608 if (iface->ofp_port == ofp_port) {
3615 /* Set Ethernet address of 'iface', if one is specified in the configuration
3618 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
3620 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
3621 struct iface *hw_addr_iface;
3623 if (strcmp(iface->type, "internal")) {
3627 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
3629 } else if (port->cfg->fake_bridge) {
3630 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
3631 find_local_hw_addr(br, ea, port, &hw_addr_iface);
3636 if (iface->ofp_port == OFPP_LOCAL) {
3637 VLOG_ERR("interface %s: ignoring mac in Interface record "
3638 "(use Bridge record to set local port's mac)",
3640 } else if (eth_addr_is_multicast(mac)) {
3641 VLOG_ERR("interface %s: cannot set MAC to multicast address",
3644 int error = netdev_set_etheraddr(iface->netdev, mac);
3646 VLOG_ERR("interface %s: setting MAC failed (%s)",
3647 iface->name, ovs_strerror(error));
3653 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
3655 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
3657 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3658 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
3659 ovsrec_interface_set_ofport(if_cfg, &port, 1);
3663 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
3664 * sets the "ofport" field to -1.
3666 * This is appropriate when 'if_cfg''s interface cannot be created or is
3667 * otherwise invalid. */
3669 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
3671 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
3672 iface_set_ofport(if_cfg, OFPP_NONE);
3673 ovsrec_interface_set_error(if_cfg, errp);
3674 ovsrec_interface_set_status(if_cfg, NULL);
3675 ovsrec_interface_set_admin_state(if_cfg, NULL);
3676 ovsrec_interface_set_duplex(if_cfg, NULL);
3677 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
3678 ovsrec_interface_set_link_state(if_cfg, NULL);
3679 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
3680 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
3681 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
3682 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
3683 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
3684 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
3685 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
3686 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
3691 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
3693 union ovsdb_atom atom;
3695 atom.integer = target;
3696 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
3700 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
3702 struct ofpbuf queues_buf;
3704 ofpbuf_init(&queues_buf, 0);
3706 if (!qos || qos->type[0] == '\0' || qos->n_queues < 1) {
3707 netdev_set_qos(iface->netdev, NULL, NULL);
3709 const struct ovsdb_datum *queues;
3710 struct netdev_queue_dump dump;
3711 unsigned int queue_id;
3712 struct smap details;
3716 /* Configure top-level Qos for 'iface'. */
3717 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
3719 /* Deconfigure queues that were deleted. */
3720 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
3722 smap_init(&details);
3723 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3724 if (!queue_ids_include(queues, queue_id)) {
3725 netdev_delete_queue(iface->netdev, queue_id);
3728 smap_destroy(&details);
3730 /* Configure queues for 'iface'. */
3732 for (i = 0; i < qos->n_queues; i++) {
3733 const struct ovsrec_queue *queue = qos->value_queues[i];
3734 unsigned int queue_id = qos->key_queues[i];
3736 if (queue_id == 0) {
3740 if (queue->n_dscp == 1) {
3741 struct ofproto_port_queue *port_queue;
3743 port_queue = ofpbuf_put_uninit(&queues_buf,
3744 sizeof *port_queue);
3745 port_queue->queue = queue_id;
3746 port_queue->dscp = queue->dscp[0];
3749 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
3752 struct smap details;
3754 smap_init(&details);
3755 netdev_set_queue(iface->netdev, 0, &details);
3756 smap_destroy(&details);
3760 if (iface->ofp_port != OFPP_NONE) {
3761 const struct ofproto_port_queue *port_queues = ofpbuf_data(&queues_buf);
3762 size_t n_queues = ofpbuf_size(&queues_buf) / sizeof *port_queues;
3764 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
3765 port_queues, n_queues);
3768 netdev_set_policing(iface->netdev,
3769 iface->cfg->ingress_policing_rate,
3770 iface->cfg->ingress_policing_burst);
3772 ofpbuf_uninit(&queues_buf);
3776 iface_configure_cfm(struct iface *iface)
3778 const struct ovsrec_interface *cfg = iface->cfg;
3779 const char *opstate_str;
3780 const char *cfm_ccm_vlan;
3781 struct cfm_settings s;
3782 struct smap netdev_args;
3784 if (!cfg->n_cfm_mpid) {
3785 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
3789 s.check_tnl_key = false;
3790 smap_init(&netdev_args);
3791 if (!netdev_get_config(iface->netdev, &netdev_args)) {
3792 const char *key = smap_get(&netdev_args, "key");
3793 const char *in_key = smap_get(&netdev_args, "in_key");
3795 s.check_tnl_key = (key && !strcmp(key, "flow"))
3796 || (in_key && !strcmp(in_key, "flow"));
3798 smap_destroy(&netdev_args);
3800 s.mpid = *cfg->cfm_mpid;
3801 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
3802 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
3803 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
3805 if (s.interval <= 0) {
3809 if (!cfm_ccm_vlan) {
3811 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
3812 s.ccm_vlan = CFM_RANDOM_VLAN;
3814 s.ccm_vlan = atoi(cfm_ccm_vlan);
3815 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
3820 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
3822 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
3824 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
3825 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
3827 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
3830 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
3831 * instead of obtaining it from the database. */
3833 iface_is_synthetic(const struct iface *iface)
3835 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
3839 iface_validate_ofport__(size_t n, int64_t *ofport)
3841 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
3842 ? u16_to_ofp(*ofport)
3847 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
3849 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
3853 iface_pick_ofport(const struct ovsrec_interface *cfg)
3855 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
3856 return (requested_ofport != OFPP_NONE
3858 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
3861 /* Port mirroring. */
3863 static struct mirror *
3864 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
3868 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
3869 if (uuid_equals(uuid, &m->uuid)) {
3877 bridge_configure_mirrors(struct bridge *br)
3879 const struct ovsdb_datum *mc;
3880 unsigned long *flood_vlans;
3881 struct mirror *m, *next;
3884 /* Get rid of deleted mirrors. */
3885 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
3886 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
3887 union ovsdb_atom atom;
3889 atom.uuid = m->uuid;
3890 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3895 /* Add new mirrors and reconfigure existing ones. */
3896 for (i = 0; i < br->cfg->n_mirrors; i++) {
3897 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
3898 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
3900 m = mirror_create(br, cfg);
3903 if (!mirror_configure(m)) {
3908 /* Update flooded vlans (for RSPAN). */
3909 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
3910 br->cfg->n_flood_vlans);
3911 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
3912 bitmap_free(flood_vlans);
3915 static struct mirror *
3916 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
3920 m = xzalloc(sizeof *m);
3921 m->uuid = cfg->header_.uuid;
3922 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
3924 m->name = xstrdup(cfg->name);
3930 mirror_destroy(struct mirror *m)
3933 struct bridge *br = m->bridge;
3936 ofproto_mirror_unregister(br->ofproto, m);
3939 hmap_remove(&br->mirrors, &m->hmap_node);
3946 mirror_collect_ports(struct mirror *m,
3947 struct ovsrec_port **in_ports, int n_in_ports,
3948 void ***out_portsp, size_t *n_out_portsp)
3950 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
3951 size_t n_out_ports = 0;
3954 for (i = 0; i < n_in_ports; i++) {
3955 const char *name = in_ports[i]->name;
3956 struct port *port = port_lookup(m->bridge, name);
3958 out_ports[n_out_ports++] = port;
3960 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
3961 "port %s", m->bridge->name, m->name, name);
3964 *out_portsp = out_ports;
3965 *n_out_portsp = n_out_ports;
3969 mirror_configure(struct mirror *m)
3971 const struct ovsrec_mirror *cfg = m->cfg;
3972 struct ofproto_mirror_settings s;
3975 if (strcmp(cfg->name, m->name)) {
3977 m->name = xstrdup(cfg->name);
3981 /* Get output port or VLAN. */
3982 if (cfg->output_port) {
3983 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
3984 if (!s.out_bundle) {
3985 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
3986 m->bridge->name, m->name);
3989 s.out_vlan = UINT16_MAX;
3991 if (cfg->output_vlan) {
3992 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
3993 "output vlan; ignoring output vlan",
3994 m->bridge->name, m->name);
3996 } else if (cfg->output_vlan) {
3997 /* The database should prevent invalid VLAN values. */
3998 s.out_bundle = NULL;
3999 s.out_vlan = *cfg->output_vlan;
4001 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4002 m->bridge->name, m->name);
4006 /* Get port selection. */
4007 if (cfg->select_all) {
4008 size_t n_ports = hmap_count(&m->bridge->ports);
4009 void **ports = xmalloc(n_ports * sizeof *ports);
4014 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4024 /* Get ports, dropping ports that don't exist.
4025 * The IDL ensures that there are no duplicates. */
4026 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4027 &s.srcs, &s.n_srcs);
4028 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4029 &s.dsts, &s.n_dsts);
4032 /* Get VLAN selection. */
4033 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4036 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4039 if (s.srcs != s.dsts) {
4048 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4050 * This is deprecated. It is only for compatibility with broken device drivers
4051 * in old versions of Linux that do not properly support VLANs when VLAN
4052 * devices are not used. When broken device drivers are no longer in
4053 * widespread use, we will delete these interfaces. */
4055 static struct ovsrec_port **recs;
4056 static size_t n_recs, allocated_recs;
4058 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4059 * splinters are reconfigured. */
4061 register_rec(struct ovsrec_port *rec)
4063 if (n_recs >= allocated_recs) {
4064 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4066 recs[n_recs++] = rec;
4069 /* Frees all of the ports registered with register_reg(). */
4071 free_registered_recs(void)
4075 for (i = 0; i < n_recs; i++) {
4076 struct ovsrec_port *port = recs[i];
4079 for (j = 0; j < port->n_interfaces; j++) {
4080 struct ovsrec_interface *iface = port->interfaces[j];
4085 smap_destroy(&port->other_config);
4086 free(port->interfaces);
4094 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4097 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4099 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4103 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4106 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4107 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4108 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4111 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4112 * use, returns NULL.
4114 * Updates 'vlan_splinters_enabled_anywhere'. */
4115 static unsigned long int *
4116 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4118 unsigned long int *splinter_vlans;
4119 struct sset splinter_ifaces;
4120 const char *real_dev_name;
4121 struct shash *real_devs;
4122 struct shash_node *node;
4126 /* Free space allocated for synthesized ports and interfaces, since we're
4127 * in the process of reconstructing all of them. */
4128 free_registered_recs();
4130 splinter_vlans = bitmap_allocate(4096);
4131 sset_init(&splinter_ifaces);
4132 vlan_splinters_enabled_anywhere = false;
4133 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4134 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4137 for (j = 0; j < br_cfg->n_ports; j++) {
4138 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4141 for (k = 0; k < port_cfg->n_interfaces; k++) {
4142 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4144 if (vlan_splinters_is_enabled(iface_cfg)) {
4145 vlan_splinters_enabled_anywhere = true;
4146 sset_add(&splinter_ifaces, iface_cfg->name);
4147 vlan_bitmap_from_array__(port_cfg->trunks,
4153 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4154 bitmap_set1(splinter_vlans, *port_cfg->tag);
4159 if (!vlan_splinters_enabled_anywhere) {
4160 free(splinter_vlans);
4161 sset_destroy(&splinter_ifaces);
4165 HMAP_FOR_EACH (br, node, &all_bridges) {
4167 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4171 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4172 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4173 * device to be created for it. */
4174 bitmap_set0(splinter_vlans, 0);
4175 bitmap_set0(splinter_vlans, 4095);
4177 /* Delete all VLAN devices that we don't need. */
4179 real_devs = vlandev_get_real_devs();
4180 SHASH_FOR_EACH (node, real_devs) {
4181 const struct vlan_real_dev *real_dev = node->data;
4182 const struct vlan_dev *vlan_dev;
4183 bool real_dev_has_splinters;
4185 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4187 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4188 if (!real_dev_has_splinters
4189 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4190 struct netdev *netdev;
4192 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4193 if (!netdev_get_in4(netdev, NULL, NULL) ||
4194 !netdev_get_in6(netdev, NULL)) {
4195 /* It has an IP address configured, so we don't own
4196 * it. Don't delete it. */
4198 vlandev_del(vlan_dev->name);
4200 netdev_close(netdev);
4207 /* Add all VLAN devices that we need. */
4208 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4211 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4212 if (!vlandev_get_name(real_dev_name, vid)) {
4213 vlandev_add(real_dev_name, vid);
4220 sset_destroy(&splinter_ifaces);
4222 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4223 free(splinter_vlans);
4226 return splinter_vlans;
4229 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4232 configure_splinter_port(struct port *port)
4234 struct ofproto *ofproto = port->bridge->ofproto;
4235 ofp_port_t realdev_ofp_port;
4236 const char *realdev_name;
4237 struct iface *vlandev, *realdev;
4239 ofproto_bundle_unregister(port->bridge->ofproto, port);
4241 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4244 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4245 realdev = iface_lookup(port->bridge, realdev_name);
4246 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4248 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4252 static struct ovsrec_port *
4253 synthesize_splinter_port(const char *real_dev_name,
4254 const char *vlan_dev_name, int vid)
4256 struct ovsrec_interface *iface;
4257 struct ovsrec_port *port;
4259 iface = xmalloc(sizeof *iface);
4260 ovsrec_interface_init(iface);
4261 iface->name = xstrdup(vlan_dev_name);
4262 iface->type = "system";
4264 port = xmalloc(sizeof *port);
4265 ovsrec_port_init(port);
4266 port->interfaces = xmemdup(&iface, sizeof iface);
4267 port->n_interfaces = 1;
4268 port->name = xstrdup(vlan_dev_name);
4269 port->vlan_mode = "splinter";
4270 port->tag = xmalloc(sizeof *port->tag);
4273 smap_add(&port->other_config, "realdev", real_dev_name);
4279 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4280 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4281 * 1-bit in the 'splinter_vlans' bitmap. */
4283 add_vlan_splinter_ports(struct bridge *br,
4284 const unsigned long int *splinter_vlans,
4285 struct shash *ports)
4289 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4290 * we're modifying 'ports'. */
4291 for (i = 0; i < br->cfg->n_ports; i++) {
4292 const char *name = br->cfg->ports[i]->name;
4293 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4296 for (j = 0; j < port_cfg->n_interfaces; j++) {
4297 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
4299 if (vlan_splinters_is_enabled(iface_cfg)) {
4300 const char *real_dev_name;
4303 real_dev_name = iface_cfg->name;
4304 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4305 const char *vlan_dev_name;
4307 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
4309 && !shash_find(ports, vlan_dev_name)) {
4310 shash_add(ports, vlan_dev_name,
4311 synthesize_splinter_port(
4312 real_dev_name, vlan_dev_name, vid));
4321 mirror_refresh_stats(struct mirror *m)
4323 struct ofproto *ofproto = m->bridge->ofproto;
4324 uint64_t tx_packets, tx_bytes;
4327 size_t stat_cnt = 0;
4329 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
4330 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
4334 if (tx_packets != UINT64_MAX) {
4335 keys[stat_cnt] = "tx_packets";
4336 values[stat_cnt] = tx_packets;
4339 if (tx_bytes != UINT64_MAX) {
4340 keys[stat_cnt] = "tx_bytes";
4341 values[stat_cnt] = tx_bytes;
4345 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);