1 /* Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 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.
22 #include "async-append.h"
26 #include "connectivity.h"
31 #include "dynamic-string.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
44 #include "ofp-print.h"
47 #include "ofproto/bond.h"
48 #include "ofproto/ofproto.h"
50 #include "poll-loop.h"
55 #include "socket-util.h"
57 #include "stream-ssl.h"
59 #include "system-stats.h"
64 #include "lib/vswitch-idl.h"
65 #include "xenserver.h"
66 #include "openvswitch/vlog.h"
67 #include "sflow_api.h"
68 #include "vlan-bitmap.h"
71 VLOG_DEFINE_THIS_MODULE(bridge);
73 COVERAGE_DEFINE(bridge_reconfigure);
76 /* These members are always valid.
78 * They are immutable: they never change between iface_create() and
80 struct ovs_list port_elem; /* Element in struct port's "ifaces" list. */
81 struct hmap_node name_node; /* In struct bridge's "iface_by_name" hmap. */
82 struct hmap_node ofp_port_node; /* In struct bridge's "ifaces" hmap. */
83 struct port *port; /* Containing port. */
84 char *name; /* Host network device name. */
85 struct netdev *netdev; /* Network device. */
86 ofp_port_t ofp_port; /* OpenFlow port number. */
89 /* These members are valid only within bridge_reconfigure(). */
90 const char *type; /* Usually same as cfg->type. */
91 const struct ovsrec_interface *cfg;
95 struct uuid uuid; /* UUID of this "mirror" record in database. */
96 struct hmap_node hmap_node; /* In struct bridge's "mirrors" hmap. */
97 struct bridge *bridge;
99 const struct ovsrec_mirror *cfg;
103 struct hmap_node hmap_node; /* Element in struct bridge's "ports" hmap. */
104 struct bridge *bridge;
107 const struct ovsrec_port *cfg;
109 /* An ordinary bridge port has 1 interface.
110 * A bridge port for bonding has at least 2 interfaces. */
111 struct ovs_list ifaces; /* List of "struct iface"s. */
115 struct hmap_node node; /* In 'all_bridges'. */
116 char *name; /* User-specified arbitrary name. */
117 char *type; /* Datapath type. */
118 uint8_t ea[ETH_ADDR_LEN]; /* Bridge Ethernet Address. */
119 uint8_t default_ea[ETH_ADDR_LEN]; /* Default MAC. */
120 const struct ovsrec_bridge *cfg;
122 /* OpenFlow switch processing. */
123 struct ofproto *ofproto; /* OpenFlow switch. */
126 struct hmap ports; /* "struct port"s indexed by name. */
127 struct hmap ifaces; /* "struct iface"s indexed by ofp_port. */
128 struct hmap iface_by_name; /* "struct iface"s indexed by name. */
130 /* Port mirroring. */
131 struct hmap mirrors; /* "struct mirror" indexed by UUID. */
134 struct hmap mappings; /* "struct" indexed by UUID */
136 /* Used during reconfiguration. */
137 struct shash wanted_ports;
139 /* Synthetic local port if necessary. */
140 struct ovsrec_port synth_local_port;
141 struct ovsrec_interface synth_local_iface;
142 struct ovsrec_interface *synth_local_ifacep;
146 struct hmap_node hmap_node; /* In struct bridge's "mappings" hmap. */
147 struct bridge *bridge;
153 /* All bridges, indexed by name. */
154 static struct hmap all_bridges = HMAP_INITIALIZER(&all_bridges);
156 /* OVSDB IDL used to obtain configuration. */
157 static struct ovsdb_idl *idl;
159 /* We want to complete daemonization, fully detaching from our parent process,
160 * only after we have completed our initial configuration, committed our state
161 * to the database, and received confirmation back from the database server
162 * that it applied the commit. This allows our parent process to know that,
163 * post-detach, ephemeral fields such as datapath-id and ofport are very likely
164 * to have already been filled in. (It is only "very likely" rather than
165 * certain because there is always a slim possibility that the transaction will
166 * fail or that some other client has added new bridges, ports, etc. while
167 * ovs-vswitchd was configuring using an old configuration.)
169 * We only need to do this once for our initial configuration at startup, so
170 * 'initial_config_done' tracks whether we've already done it. While we are
171 * waiting for a response to our commit, 'daemonize_txn' tracks the transaction
172 * itself and is otherwise NULL. */
173 static bool initial_config_done;
174 static struct ovsdb_idl_txn *daemonize_txn;
176 /* Most recently processed IDL sequence number. */
177 static unsigned int idl_seqno;
179 /* Track changes to port connectivity. */
180 static uint64_t connectivity_seqno = LLONG_MIN;
182 /* Status update to database.
184 * Some information in the database must be kept as up-to-date as possible to
185 * allow controllers to respond rapidly to network outages. Those status are
186 * updated via the 'status_txn'.
188 * We use the global connectivity sequence number to detect the status change.
189 * Also, to prevent the status update from sending too much to the database,
190 * we check the return status of each update transaction and do not start new
191 * update if the previous transaction status is 'TXN_INCOMPLETE'.
193 * 'statux_txn' is NULL if there is no ongoing status update.
195 * If the previous database transaction was failed (is not 'TXN_SUCCESS',
196 * 'TXN_UNCHANGED' or 'TXN_INCOMPLETE'), 'status_txn_try_again' is set to true,
197 * which will cause the main thread wake up soon and retry the status update.
199 static struct ovsdb_idl_txn *status_txn;
200 static bool status_txn_try_again;
202 /* When the status update transaction returns 'TXN_INCOMPLETE', should register a
203 * timeout in 'STATUS_CHECK_AGAIN_MSEC' to check again. */
204 #define STATUS_CHECK_AGAIN_MSEC 100
206 /* Statistics update to database. */
207 static struct ovsdb_idl_txn *stats_txn;
209 /* Each time this timer expires, the bridge fetches interface and mirror
210 * statistics and pushes them into the database. */
211 static int stats_timer_interval;
212 static long long int stats_timer = LLONG_MIN;
214 /* Each time this timer expires, the bridge fetches the list of port/VLAN
215 * membership that has been modified by the AA.
217 #define AA_REFRESH_INTERVAL (1000) /* In milliseconds. */
218 static long long int aa_refresh_timer = LLONG_MIN;
220 /* In some datapaths, creating and destroying OpenFlow ports can be extremely
221 * expensive. This can cause bridge_reconfigure() to take a long time during
222 * which no other work can be done. To deal with this problem, we limit port
223 * adds and deletions to a window of OFP_PORT_ACTION_WINDOW milliseconds per
224 * call to bridge_reconfigure(). If there is more work to do after the limit
225 * is reached, 'need_reconfigure', is flagged and it's done on the next loop.
226 * This allows the rest of the code to catch up on important things like
227 * forwarding packets. */
228 #define OFP_PORT_ACTION_WINDOW 10
230 static void add_del_bridges(const struct ovsrec_open_vswitch *);
231 static void bridge_run__(void);
232 static void bridge_create(const struct ovsrec_bridge *);
233 static void bridge_destroy(struct bridge *);
234 static struct bridge *bridge_lookup(const char *name);
235 static unixctl_cb_func bridge_unixctl_dump_flows;
236 static unixctl_cb_func bridge_unixctl_reconnect;
237 static size_t bridge_get_controllers(const struct bridge *br,
238 struct ovsrec_controller ***controllersp);
239 static void bridge_collect_wanted_ports(struct bridge *,
240 const unsigned long *splinter_vlans,
241 struct shash *wanted_ports);
242 static void bridge_delete_ofprotos(void);
243 static void bridge_delete_or_reconfigure_ports(struct bridge *);
244 static void bridge_del_ports(struct bridge *,
245 const struct shash *wanted_ports);
246 static void bridge_add_ports(struct bridge *,
247 const struct shash *wanted_ports);
249 static void bridge_configure_datapath_id(struct bridge *);
250 static void bridge_configure_netflow(struct bridge *);
251 static void bridge_configure_forward_bpdu(struct bridge *);
252 static void bridge_configure_mac_table(struct bridge *);
253 static void bridge_configure_mcast_snooping(struct bridge *);
254 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
255 static void bridge_configure_ipfix(struct bridge *);
256 static void bridge_configure_spanning_tree(struct bridge *);
257 static void bridge_configure_tables(struct bridge *);
258 static void bridge_configure_dp_desc(struct bridge *);
259 static void bridge_configure_aa(struct bridge *);
260 static void bridge_aa_refresh_queued(struct bridge *);
261 static bool bridge_aa_need_refresh(struct bridge *);
262 static void bridge_configure_remotes(struct bridge *,
263 const struct sockaddr_in *managers,
265 static void bridge_pick_local_hw_addr(struct bridge *,
266 uint8_t ea[ETH_ADDR_LEN],
267 struct iface **hw_addr_iface);
268 static uint64_t bridge_pick_datapath_id(struct bridge *,
269 const uint8_t bridge_ea[ETH_ADDR_LEN],
270 struct iface *hw_addr_iface);
271 static uint64_t dpid_from_hash(const void *, size_t nbytes);
272 static bool bridge_has_bond_fake_iface(const struct bridge *,
274 static bool port_is_bond_fake_iface(const struct port *);
276 static unixctl_cb_func qos_unixctl_show;
278 static struct port *port_create(struct bridge *, const struct ovsrec_port *);
279 static void port_del_ifaces(struct port *);
280 static void port_destroy(struct port *);
281 static struct port *port_lookup(const struct bridge *, const char *name);
282 static void port_configure(struct port *);
283 static struct lacp_settings *port_configure_lacp(struct port *,
284 struct lacp_settings *);
285 static void port_configure_bond(struct port *, struct bond_settings *);
286 static bool port_is_synthetic(const struct port *);
288 static void reconfigure_system_stats(const struct ovsrec_open_vswitch *);
289 static void run_system_stats(void);
291 static void bridge_configure_mirrors(struct bridge *);
292 static struct mirror *mirror_create(struct bridge *,
293 const struct ovsrec_mirror *);
294 static void mirror_destroy(struct mirror *);
295 static bool mirror_configure(struct mirror *);
296 static void mirror_refresh_stats(struct mirror *);
298 static void iface_configure_lacp(struct iface *, struct lacp_slave_settings *);
299 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
300 const struct ovsrec_port *);
301 static bool iface_is_internal(const struct ovsrec_interface *iface,
302 const struct ovsrec_bridge *br);
303 static const char *iface_get_type(const struct ovsrec_interface *,
304 const struct ovsrec_bridge *);
305 static void iface_destroy(struct iface *);
306 static void iface_destroy__(struct iface *);
307 static struct iface *iface_lookup(const struct bridge *, const char *name);
308 static struct iface *iface_find(const char *name);
309 static struct iface *iface_from_ofp_port(const struct bridge *,
310 ofp_port_t ofp_port);
311 static void iface_set_mac(const struct bridge *, const struct port *, struct iface *);
312 static void iface_set_ofport(const struct ovsrec_interface *, ofp_port_t ofport);
313 static void iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp);
314 static void iface_configure_qos(struct iface *, const struct ovsrec_qos *);
315 static void iface_configure_cfm(struct iface *);
316 static void iface_refresh_cfm_stats(struct iface *);
317 static void iface_refresh_stats(struct iface *);
318 static void iface_refresh_netdev_status(struct iface *);
319 static void iface_refresh_ofproto_status(struct iface *);
320 static bool iface_is_synthetic(const struct iface *);
321 static ofp_port_t iface_get_requested_ofp_port(
322 const struct ovsrec_interface *);
323 static ofp_port_t iface_pick_ofport(const struct ovsrec_interface *);
326 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
328 * This is deprecated. It is only for compatibility with broken device drivers
329 * in old versions of Linux that do not properly support VLANs when VLAN
330 * devices are not used. When broken device drivers are no longer in
331 * widespread use, we will delete these interfaces. */
333 /* True if VLAN splinters are enabled on any interface, false otherwise.*/
334 static bool vlan_splinters_enabled_anywhere;
336 static bool vlan_splinters_is_enabled(const struct ovsrec_interface *);
337 static unsigned long int *collect_splinter_vlans(
338 const struct ovsrec_open_vswitch *);
339 static void configure_splinter_port(struct port *);
340 static void add_vlan_splinter_ports(struct bridge *,
341 const unsigned long int *splinter_vlans,
342 struct shash *ports);
344 static void discover_types(const struct ovsrec_open_vswitch *cfg);
347 bridge_init_ofproto(const struct ovsrec_open_vswitch *cfg)
349 struct shash iface_hints;
350 static bool initialized = false;
357 shash_init(&iface_hints);
360 for (i = 0; i < cfg->n_bridges; i++) {
361 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
364 for (j = 0; j < br_cfg->n_ports; j++) {
365 struct ovsrec_port *port_cfg = br_cfg->ports[j];
368 for (k = 0; k < port_cfg->n_interfaces; k++) {
369 struct ovsrec_interface *if_cfg = port_cfg->interfaces[k];
370 struct iface_hint *iface_hint;
372 iface_hint = xmalloc(sizeof *iface_hint);
373 iface_hint->br_name = br_cfg->name;
374 iface_hint->br_type = br_cfg->datapath_type;
375 iface_hint->ofp_port = iface_pick_ofport(if_cfg);
377 shash_add(&iface_hints, if_cfg->name, iface_hint);
383 ofproto_init(&iface_hints);
385 shash_destroy_free_data(&iface_hints);
389 /* Public functions. */
391 /* Initializes the bridge module, configuring it to obtain its configuration
392 * from an OVSDB server accessed over 'remote', which should be a string in a
393 * form acceptable to ovsdb_idl_create(). */
395 bridge_init(const char *remote)
397 /* Create connection to database. */
398 idl = ovsdb_idl_create(remote, &ovsrec_idl_class, true, true);
399 idl_seqno = ovsdb_idl_get_seqno(idl);
400 ovsdb_idl_set_lock(idl, "ovs_vswitchd");
401 ovsdb_idl_verify_write_only(idl);
403 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_cur_cfg);
404 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_statistics);
405 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_datapath_types);
406 ovsdb_idl_omit_alert(idl, &ovsrec_open_vswitch_col_iface_types);
407 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_external_ids);
408 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_ovs_version);
409 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_db_version);
410 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_type);
411 ovsdb_idl_omit(idl, &ovsrec_open_vswitch_col_system_version);
413 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_id);
414 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_datapath_version);
415 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_status);
416 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_status);
417 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_stp_enable);
418 ovsdb_idl_omit_alert(idl, &ovsrec_bridge_col_rstp_enable);
419 ovsdb_idl_omit(idl, &ovsrec_bridge_col_external_ids);
421 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_status);
422 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_status);
423 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_rstp_statistics);
424 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_statistics);
425 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_bond_active_slave);
426 ovsdb_idl_omit(idl, &ovsrec_port_col_external_ids);
427 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_trunks);
428 ovsdb_idl_omit_alert(idl, &ovsrec_port_col_vlan_mode);
429 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_admin_state);
430 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_duplex);
431 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_speed);
432 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_state);
433 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_link_resets);
434 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mac_in_use);
435 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ifindex);
436 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_mtu);
437 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_ofport);
438 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_statistics);
439 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_status);
440 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault);
441 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_fault_status);
442 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_mpids);
443 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_flap_count);
444 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_health);
445 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_cfm_remote_opstate);
446 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_bfd_status);
447 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_lacp_current);
448 ovsdb_idl_omit_alert(idl, &ovsrec_interface_col_error);
449 ovsdb_idl_omit(idl, &ovsrec_interface_col_external_ids);
451 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_is_connected);
452 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_role);
453 ovsdb_idl_omit_alert(idl, &ovsrec_controller_col_status);
454 ovsdb_idl_omit(idl, &ovsrec_controller_col_external_ids);
456 ovsdb_idl_omit(idl, &ovsrec_qos_col_external_ids);
458 ovsdb_idl_omit(idl, &ovsrec_queue_col_external_ids);
460 ovsdb_idl_omit(idl, &ovsrec_mirror_col_external_ids);
461 ovsdb_idl_omit_alert(idl, &ovsrec_mirror_col_statistics);
463 ovsdb_idl_omit(idl, &ovsrec_netflow_col_external_ids);
464 ovsdb_idl_omit(idl, &ovsrec_sflow_col_external_ids);
465 ovsdb_idl_omit(idl, &ovsrec_ipfix_col_external_ids);
466 ovsdb_idl_omit(idl, &ovsrec_flow_sample_collector_set_col_external_ids);
468 ovsdb_idl_omit(idl, &ovsrec_manager_col_external_ids);
469 ovsdb_idl_omit(idl, &ovsrec_manager_col_inactivity_probe);
470 ovsdb_idl_omit(idl, &ovsrec_manager_col_is_connected);
471 ovsdb_idl_omit(idl, &ovsrec_manager_col_max_backoff);
472 ovsdb_idl_omit(idl, &ovsrec_manager_col_status);
474 ovsdb_idl_omit(idl, &ovsrec_ssl_col_external_ids);
476 /* Register unixctl commands. */
477 unixctl_command_register("qos/show", "interface", 1, 1,
478 qos_unixctl_show, NULL);
479 unixctl_command_register("bridge/dump-flows", "bridge", 1, 1,
480 bridge_unixctl_dump_flows, NULL);
481 unixctl_command_register("bridge/reconnect", "[bridge]", 0, 1,
482 bridge_unixctl_reconnect, NULL);
495 struct bridge *br, *next_br;
497 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
500 ovsdb_idl_destroy(idl);
503 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
504 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
505 * responsible for freeing '*managersp' (with free()).
507 * You may be asking yourself "why does ovs-vswitchd care?", because
508 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
509 * should not be and in fact is not directly involved in that. But
510 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
511 * it has to tell in-band control where the managers are to enable that.
512 * (Thus, only managers connected in-band are collected.)
515 collect_in_band_managers(const struct ovsrec_open_vswitch *ovs_cfg,
516 struct sockaddr_in **managersp, size_t *n_managersp)
518 struct sockaddr_in *managers = NULL;
519 size_t n_managers = 0;
523 /* Collect all of the potential targets from the "targets" columns of the
524 * rows pointed to by "manager_options", excluding any that are
527 for (i = 0; i < ovs_cfg->n_manager_options; i++) {
528 struct ovsrec_manager *m = ovs_cfg->manager_options[i];
530 if (m->connection_mode && !strcmp(m->connection_mode, "out-of-band")) {
531 sset_find_and_delete(&targets, m->target);
533 sset_add(&targets, m->target);
537 /* Now extract the targets' IP addresses. */
538 if (!sset_is_empty(&targets)) {
541 managers = xmalloc(sset_count(&targets) * sizeof *managers);
542 SSET_FOR_EACH (target, &targets) {
544 struct sockaddr_storage ss;
545 struct sockaddr_in in;
548 if (stream_parse_target_with_default_port(target, OVSDB_PORT,
550 && sa.ss.ss_family == AF_INET) {
551 managers[n_managers++] = sa.in;
555 sset_destroy(&targets);
557 *managersp = managers;
558 *n_managersp = n_managers;
562 bridge_reconfigure(const struct ovsrec_open_vswitch *ovs_cfg)
564 unsigned long int *splinter_vlans;
565 struct sockaddr_in *managers;
566 struct bridge *br, *next;
567 int sflow_bridge_number;
570 COVERAGE_INC(bridge_reconfigure);
572 ofproto_set_flow_limit(smap_get_int(&ovs_cfg->other_config, "flow-limit",
573 OFPROTO_FLOW_LIMIT_DEFAULT));
574 ofproto_set_max_idle(smap_get_int(&ovs_cfg->other_config, "max-idle",
575 OFPROTO_MAX_IDLE_DEFAULT));
576 ofproto_set_n_dpdk_rxqs(smap_get_int(&ovs_cfg->other_config,
578 ofproto_set_cpu_mask(smap_get(&ovs_cfg->other_config, "pmd-cpu-mask"));
581 smap_get_int(&ovs_cfg->other_config, "n-handler-threads", 0),
582 smap_get_int(&ovs_cfg->other_config, "n-revalidator-threads", 0));
585 discover_types(ovs_cfg);
588 /* Destroy "struct bridge"s, "struct port"s, and "struct iface"s according
589 * to 'ovs_cfg', with only very minimal configuration otherwise.
591 * This is mostly an update to bridge data structures. Nothing is pushed
592 * down to ofproto or lower layers. */
593 add_del_bridges(ovs_cfg);
594 splinter_vlans = collect_splinter_vlans(ovs_cfg);
595 HMAP_FOR_EACH (br, node, &all_bridges) {
596 bridge_collect_wanted_ports(br, splinter_vlans, &br->wanted_ports);
597 bridge_del_ports(br, &br->wanted_ports);
599 free(splinter_vlans);
601 /* Start pushing configuration changes down to the ofproto layer:
603 * - Delete ofprotos that are no longer configured.
605 * - Delete ports that are no longer configured.
607 * - Reconfigure existing ports to their desired configurations, or
608 * delete them if not possible.
610 * We have to do all the deletions before we can do any additions, because
611 * the ports to be added might require resources that will be freed up by
612 * deletions (they might especially overlap in name). */
613 bridge_delete_ofprotos();
614 HMAP_FOR_EACH (br, node, &all_bridges) {
616 bridge_delete_or_reconfigure_ports(br);
620 /* Finish pushing configuration changes to the ofproto layer:
622 * - Create ofprotos that are missing.
624 * - Add ports that are missing. */
625 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
629 error = ofproto_create(br->name, br->type, &br->ofproto);
631 VLOG_ERR("failed to create bridge %s: %s", br->name,
632 ovs_strerror(error));
633 shash_destroy(&br->wanted_ports);
636 /* Trigger storing datapath version. */
637 seq_change(connectivity_seq_get());
641 HMAP_FOR_EACH (br, node, &all_bridges) {
642 bridge_add_ports(br, &br->wanted_ports);
643 shash_destroy(&br->wanted_ports);
646 reconfigure_system_stats(ovs_cfg);
648 /* Complete the configuration. */
649 sflow_bridge_number = 0;
650 collect_in_band_managers(ovs_cfg, &managers, &n_managers);
651 HMAP_FOR_EACH (br, node, &all_bridges) {
654 /* We need the datapath ID early to allow LACP ports to use it as the
655 * default system ID. */
656 bridge_configure_datapath_id(br);
658 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
661 port_configure(port);
663 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
664 iface_set_ofport(iface->cfg, iface->ofp_port);
665 /* Clear eventual previous errors */
666 ovsrec_interface_set_error(iface->cfg, NULL);
667 iface_configure_cfm(iface);
668 iface_configure_qos(iface, port->cfg->qos);
669 iface_set_mac(br, port, iface);
670 ofproto_port_set_bfd(br->ofproto, iface->ofp_port,
672 ofproto_port_set_lldp(br->ofproto, iface->ofp_port,
676 bridge_configure_mirrors(br);
677 bridge_configure_forward_bpdu(br);
678 bridge_configure_mac_table(br);
679 bridge_configure_mcast_snooping(br);
680 bridge_configure_remotes(br, managers, n_managers);
681 bridge_configure_netflow(br);
682 bridge_configure_sflow(br, &sflow_bridge_number);
683 bridge_configure_ipfix(br);
684 bridge_configure_spanning_tree(br);
685 bridge_configure_tables(br);
686 bridge_configure_dp_desc(br);
687 bridge_configure_aa(br);
691 /* The ofproto-dpif provider does some final reconfiguration in its
692 * ->type_run() function. We have to call it before notifying the database
693 * client that reconfiguration is complete, otherwise there is a very
694 * narrow race window in which e.g. ofproto/trace will not recognize the
695 * new configuration (sometimes this causes unit test failures). */
699 /* Delete ofprotos which aren't configured or have the wrong type. Create
700 * ofprotos which don't exist but need to. */
702 bridge_delete_ofprotos(void)
709 /* Delete ofprotos with no bridge or with the wrong type. */
712 ofproto_enumerate_types(&types);
713 SSET_FOR_EACH (type, &types) {
716 ofproto_enumerate_names(type, &names);
717 SSET_FOR_EACH (name, &names) {
718 br = bridge_lookup(name);
719 if (!br || strcmp(type, br->type)) {
720 ofproto_delete(name, type);
724 sset_destroy(&names);
725 sset_destroy(&types);
729 add_ofp_port(ofp_port_t port, ofp_port_t *ports, size_t *n, size_t *allocated)
731 if (*n >= *allocated) {
732 ports = x2nrealloc(ports, allocated, sizeof *ports);
734 ports[(*n)++] = port;
739 bridge_delete_or_reconfigure_ports(struct bridge *br)
741 struct ofproto_port ofproto_port;
742 struct ofproto_port_dump dump;
744 struct sset ofproto_ports;
745 struct port *port, *port_next;
747 /* List of "ofp_port"s to delete. We make a list instead of deleting them
748 * right away because ofproto implementations aren't necessarily able to
749 * iterate through a changing list of ports in an entirely robust way. */
756 sset_init(&ofproto_ports);
758 /* Main task: Iterate over the ports in 'br->ofproto' and remove the ports
759 * that are not configured in the database. (This commonly happens when
760 * ports have been deleted, e.g. with "ovs-vsctl del-port".)
762 * Side tasks: Reconfigure the ports that are still in 'br'. Delete ports
763 * that have the wrong OpenFlow port number (and arrange to add them back
764 * with the correct OpenFlow port number). */
765 OFPROTO_PORT_FOR_EACH (&ofproto_port, &dump, br->ofproto) {
766 ofp_port_t requested_ofp_port;
769 sset_add(&ofproto_ports, ofproto_port.name);
771 iface = iface_lookup(br, ofproto_port.name);
773 /* No such iface is configured, so we should delete this
776 * As a corner case exception, keep the port if it's a bond fake
778 if (bridge_has_bond_fake_iface(br, ofproto_port.name)
779 && !strcmp(ofproto_port.type, "internal")) {
785 if (strcmp(ofproto_port.type, iface->type)
786 || netdev_set_config(iface->netdev, &iface->cfg->options, NULL)) {
787 /* The interface is the wrong type or can't be configured.
792 /* If the requested OpenFlow port for 'iface' changed, and it's not
793 * already the correct port, then we might want to temporarily delete
794 * this interface, so we can add it back again with the new OpenFlow
796 requested_ofp_port = iface_get_requested_ofp_port(iface->cfg);
797 if (iface->ofp_port != OFPP_LOCAL &&
798 requested_ofp_port != OFPP_NONE &&
799 requested_ofp_port != iface->ofp_port) {
800 ofp_port_t victim_request;
801 struct iface *victim;
803 /* Check for an existing OpenFlow port currently occupying
804 * 'iface''s requested port number. If there isn't one, then
805 * delete this port. Otherwise we need to consider further. */
806 victim = iface_from_ofp_port(br, requested_ofp_port);
811 /* 'victim' is a port currently using 'iface''s requested port
812 * number. Unless 'victim' specifically requested that port
813 * number, too, then we can delete both 'iface' and 'victim'
814 * temporarily. (We'll add both of them back again later with new
815 * OpenFlow port numbers.)
817 * If 'victim' did request port number 'requested_ofp_port', just
818 * like 'iface', then that's a configuration inconsistency that we
819 * can't resolve. We might as well let it keep its current port
821 victim_request = iface_get_requested_ofp_port(victim->cfg);
822 if (victim_request != requested_ofp_port) {
823 del = add_ofp_port(victim->ofp_port, del, &n, &allocated);
824 iface_destroy(victim);
833 iface_destroy(iface);
834 del = add_ofp_port(ofproto_port.ofp_port, del, &n, &allocated);
836 for (i = 0; i < n; i++) {
837 ofproto_port_del(br->ofproto, del[i]);
841 /* Iterate over this module's idea of interfaces in 'br'. Remove any ports
842 * that we didn't see when we iterated through the datapath, i.e. ports
843 * that disappeared underneath use. This is an unusual situation, but it
844 * can happen in some cases:
846 * - An admin runs a command like "ovs-dpctl del-port" (which is a bad
847 * idea but could happen).
849 * - The port represented a device that disappeared, e.g. a tuntap
850 * device destroyed via "tunctl -d", a physical Ethernet device
851 * whose module was just unloaded via "rmmod", or a virtual NIC for a
852 * VM whose VM was just terminated. */
853 HMAP_FOR_EACH_SAFE (port, port_next, hmap_node, &br->ports) {
854 struct iface *iface, *iface_next;
856 LIST_FOR_EACH_SAFE (iface, iface_next, port_elem, &port->ifaces) {
857 if (!sset_contains(&ofproto_ports, iface->name)) {
858 iface_destroy__(iface);
862 if (list_is_empty(&port->ifaces)) {
866 sset_destroy(&ofproto_ports);
870 bridge_add_ports__(struct bridge *br, const struct shash *wanted_ports,
871 bool with_requested_port)
873 struct shash_node *port_node;
875 SHASH_FOR_EACH (port_node, wanted_ports) {
876 const struct ovsrec_port *port_cfg = port_node->data;
879 for (i = 0; i < port_cfg->n_interfaces; i++) {
880 const struct ovsrec_interface *iface_cfg = port_cfg->interfaces[i];
881 ofp_port_t requested_ofp_port;
883 requested_ofp_port = iface_get_requested_ofp_port(iface_cfg);
884 if ((requested_ofp_port != OFPP_NONE) == with_requested_port) {
885 struct iface *iface = iface_lookup(br, iface_cfg->name);
888 iface_create(br, iface_cfg, port_cfg);
896 bridge_add_ports(struct bridge *br, const struct shash *wanted_ports)
898 /* First add interfaces that request a particular port number. */
899 bridge_add_ports__(br, wanted_ports, true);
901 /* Then add interfaces that want automatic port number assignment.
902 * We add these afterward to avoid accidentally taking a specifically
903 * requested port number. */
904 bridge_add_ports__(br, wanted_ports, false);
908 port_configure(struct port *port)
910 const struct ovsrec_port *cfg = port->cfg;
911 struct bond_settings bond_settings;
912 struct lacp_settings lacp_settings;
913 struct ofproto_bundle_settings s;
916 if (cfg->vlan_mode && !strcmp(cfg->vlan_mode, "splinter")) {
917 configure_splinter_port(port);
926 s.slaves = xmalloc(list_size(&port->ifaces) * sizeof *s.slaves);
927 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
928 s.slaves[s.n_slaves++] = iface->ofp_port;
933 if (cfg->tag && *cfg->tag >= 0 && *cfg->tag <= 4095) {
937 /* Get VLAN trunks. */
940 s.trunks = vlan_bitmap_from_array(cfg->trunks, cfg->n_trunks);
944 if (cfg->vlan_mode) {
945 if (!strcmp(cfg->vlan_mode, "access")) {
946 s.vlan_mode = PORT_VLAN_ACCESS;
947 } else if (!strcmp(cfg->vlan_mode, "trunk")) {
948 s.vlan_mode = PORT_VLAN_TRUNK;
949 } else if (!strcmp(cfg->vlan_mode, "native-tagged")) {
950 s.vlan_mode = PORT_VLAN_NATIVE_TAGGED;
951 } else if (!strcmp(cfg->vlan_mode, "native-untagged")) {
952 s.vlan_mode = PORT_VLAN_NATIVE_UNTAGGED;
954 /* This "can't happen" because ovsdb-server should prevent it. */
955 VLOG_WARN("port %s: unknown VLAN mode %s, falling "
956 "back to trunk mode", port->name, cfg->vlan_mode);
957 s.vlan_mode = PORT_VLAN_TRUNK;
961 s.vlan_mode = PORT_VLAN_ACCESS;
963 VLOG_WARN("port %s: ignoring trunks in favor of implicit vlan",
967 s.vlan_mode = PORT_VLAN_TRUNK;
970 s.use_priority_tags = smap_get_bool(&cfg->other_config, "priority-tags",
973 /* Get LACP settings. */
974 s.lacp = port_configure_lacp(port, &lacp_settings);
978 s.lacp_slaves = xmalloc(s.n_slaves * sizeof *s.lacp_slaves);
979 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
980 iface_configure_lacp(iface, &s.lacp_slaves[i++]);
983 s.lacp_slaves = NULL;
986 /* Get bond settings. */
987 if (s.n_slaves > 1) {
988 s.bond = &bond_settings;
989 port_configure_bond(port, &bond_settings);
992 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
993 netdev_set_miimon_interval(iface->netdev, 0);
998 ofproto_bundle_register(port->bridge->ofproto, port, &s);
1003 free(s.lacp_slaves);
1006 /* Pick local port hardware address and datapath ID for 'br'. */
1008 bridge_configure_datapath_id(struct bridge *br)
1010 uint8_t ea[ETH_ADDR_LEN];
1012 struct iface *local_iface;
1013 struct iface *hw_addr_iface;
1016 bridge_pick_local_hw_addr(br, ea, &hw_addr_iface);
1017 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
1019 int error = netdev_set_etheraddr(local_iface->netdev, ea);
1021 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1022 VLOG_ERR_RL(&rl, "bridge %s: failed to set bridge "
1023 "Ethernet address: %s",
1024 br->name, ovs_strerror(error));
1027 memcpy(br->ea, ea, ETH_ADDR_LEN);
1029 dpid = bridge_pick_datapath_id(br, ea, hw_addr_iface);
1030 if (dpid != ofproto_get_datapath_id(br->ofproto)) {
1031 VLOG_INFO("bridge %s: using datapath ID %016"PRIx64, br->name, dpid);
1032 ofproto_set_datapath_id(br->ofproto, dpid);
1035 dpid_string = xasprintf("%016"PRIx64, dpid);
1036 ovsrec_bridge_set_datapath_id(br->cfg, dpid_string);
1040 /* Returns a bitmap of "enum ofputil_protocol"s that are allowed for use with
1043 bridge_get_allowed_versions(struct bridge *br)
1045 if (!br->cfg->n_protocols)
1048 return ofputil_versions_from_strings(br->cfg->protocols,
1049 br->cfg->n_protocols);
1052 /* Set NetFlow configuration on 'br'. */
1054 bridge_configure_netflow(struct bridge *br)
1056 struct ovsrec_netflow *cfg = br->cfg->netflow;
1057 struct netflow_options opts;
1060 ofproto_set_netflow(br->ofproto, NULL);
1064 memset(&opts, 0, sizeof opts);
1066 /* Get default NetFlow configuration from datapath.
1067 * Apply overrides from 'cfg'. */
1068 ofproto_get_netflow_ids(br->ofproto, &opts.engine_type, &opts.engine_id);
1069 if (cfg->engine_type) {
1070 opts.engine_type = *cfg->engine_type;
1072 if (cfg->engine_id) {
1073 opts.engine_id = *cfg->engine_id;
1076 /* Configure active timeout interval. */
1077 opts.active_timeout = cfg->active_timeout;
1078 if (!opts.active_timeout) {
1079 opts.active_timeout = -1;
1080 } else if (opts.active_timeout < 0) {
1081 VLOG_WARN("bridge %s: active timeout interval set to negative "
1082 "value, using default instead (%d seconds)", br->name,
1083 NF_ACTIVE_TIMEOUT_DEFAULT);
1084 opts.active_timeout = -1;
1087 /* Add engine ID to interface number to disambiguate bridgs? */
1088 opts.add_id_to_iface = cfg->add_id_to_interface;
1089 if (opts.add_id_to_iface) {
1090 if (opts.engine_id > 0x7f) {
1091 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
1092 "another vswitch, choose an engine id less than 128",
1095 if (hmap_count(&br->ports) > 508) {
1096 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
1097 "another port when more than 508 ports are used",
1103 sset_init(&opts.collectors);
1104 sset_add_array(&opts.collectors, cfg->targets, cfg->n_targets);
1107 if (ofproto_set_netflow(br->ofproto, &opts)) {
1108 VLOG_ERR("bridge %s: problem setting netflow collectors", br->name);
1110 sset_destroy(&opts.collectors);
1113 /* Set sFlow configuration on 'br'. */
1115 bridge_configure_sflow(struct bridge *br, int *sflow_bridge_number)
1117 const struct ovsrec_sflow *cfg = br->cfg->sflow;
1118 struct ovsrec_controller **controllers;
1119 struct ofproto_sflow_options oso;
1120 size_t n_controllers;
1124 ofproto_set_sflow(br->ofproto, NULL);
1128 memset(&oso, 0, sizeof oso);
1130 sset_init(&oso.targets);
1131 sset_add_array(&oso.targets, cfg->targets, cfg->n_targets);
1133 oso.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1134 if (cfg->sampling) {
1135 oso.sampling_rate = *cfg->sampling;
1138 oso.polling_interval = SFL_DEFAULT_POLLING_INTERVAL;
1140 oso.polling_interval = *cfg->polling;
1143 oso.header_len = SFL_DEFAULT_HEADER_SIZE;
1145 oso.header_len = *cfg->header;
1148 oso.sub_id = (*sflow_bridge_number)++;
1149 oso.agent_device = cfg->agent;
1151 oso.control_ip = NULL;
1152 n_controllers = bridge_get_controllers(br, &controllers);
1153 for (i = 0; i < n_controllers; i++) {
1154 if (controllers[i]->local_ip) {
1155 oso.control_ip = controllers[i]->local_ip;
1159 ofproto_set_sflow(br->ofproto, &oso);
1161 sset_destroy(&oso.targets);
1164 /* Returns whether a IPFIX row is valid. */
1166 ovsrec_ipfix_is_valid(const struct ovsrec_ipfix *ipfix)
1168 return ipfix && ipfix->n_targets > 0;
1171 /* Returns whether a Flow_Sample_Collector_Set row is valid. */
1173 ovsrec_fscs_is_valid(const struct ovsrec_flow_sample_collector_set *fscs,
1174 const struct bridge *br)
1176 return ovsrec_ipfix_is_valid(fscs->ipfix) && fscs->bridge == br->cfg;
1179 /* Set IPFIX configuration on 'br'. */
1181 bridge_configure_ipfix(struct bridge *br)
1183 const struct ovsrec_ipfix *be_cfg = br->cfg->ipfix;
1184 bool valid_be_cfg = ovsrec_ipfix_is_valid(be_cfg);
1185 const struct ovsrec_flow_sample_collector_set *fe_cfg;
1186 struct ofproto_ipfix_bridge_exporter_options be_opts;
1187 struct ofproto_ipfix_flow_exporter_options *fe_opts = NULL;
1188 size_t n_fe_opts = 0;
1190 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1191 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1196 if (!valid_be_cfg && n_fe_opts == 0) {
1197 ofproto_set_ipfix(br->ofproto, NULL, NULL, 0);
1202 memset(&be_opts, 0, sizeof be_opts);
1204 sset_init(&be_opts.targets);
1205 sset_add_array(&be_opts.targets, be_cfg->targets, be_cfg->n_targets);
1207 if (be_cfg->sampling) {
1208 be_opts.sampling_rate = *be_cfg->sampling;
1210 be_opts.sampling_rate = SFL_DEFAULT_SAMPLING_RATE;
1212 if (be_cfg->obs_domain_id) {
1213 be_opts.obs_domain_id = *be_cfg->obs_domain_id;
1215 if (be_cfg->obs_point_id) {
1216 be_opts.obs_point_id = *be_cfg->obs_point_id;
1218 if (be_cfg->cache_active_timeout) {
1219 be_opts.cache_active_timeout = *be_cfg->cache_active_timeout;
1221 if (be_cfg->cache_max_flows) {
1222 be_opts.cache_max_flows = *be_cfg->cache_max_flows;
1225 be_opts.enable_tunnel_sampling = smap_get_bool(&be_cfg->other_config,
1226 "enable-tunnel-sampling", true);
1228 be_opts.enable_input_sampling = !smap_get_bool(&be_cfg->other_config,
1229 "enable-input-sampling", false);
1231 be_opts.enable_output_sampling = !smap_get_bool(&be_cfg->other_config,
1232 "enable-output-sampling", false);
1235 if (n_fe_opts > 0) {
1236 struct ofproto_ipfix_flow_exporter_options *opts;
1237 fe_opts = xcalloc(n_fe_opts, sizeof *fe_opts);
1239 OVSREC_FLOW_SAMPLE_COLLECTOR_SET_FOR_EACH(fe_cfg, idl) {
1240 if (ovsrec_fscs_is_valid(fe_cfg, br)) {
1241 opts->collector_set_id = fe_cfg->id;
1242 sset_init(&opts->targets);
1243 sset_add_array(&opts->targets, fe_cfg->ipfix->targets,
1244 fe_cfg->ipfix->n_targets);
1245 opts->cache_active_timeout = fe_cfg->ipfix->cache_active_timeout
1246 ? *fe_cfg->ipfix->cache_active_timeout : 0;
1247 opts->cache_max_flows = fe_cfg->ipfix->cache_max_flows
1248 ? *fe_cfg->ipfix->cache_max_flows : 0;
1254 ofproto_set_ipfix(br->ofproto, valid_be_cfg ? &be_opts : NULL, fe_opts,
1258 sset_destroy(&be_opts.targets);
1261 if (n_fe_opts > 0) {
1262 struct ofproto_ipfix_flow_exporter_options *opts = fe_opts;
1264 for (i = 0; i < n_fe_opts; i++) {
1265 sset_destroy(&opts->targets);
1273 port_configure_stp(const struct ofproto *ofproto, struct port *port,
1274 struct ofproto_port_stp_settings *port_s,
1275 int *port_num_counter, unsigned long *port_num_bitmap)
1277 const char *config_str;
1278 struct iface *iface;
1280 if (!smap_get_bool(&port->cfg->other_config, "stp-enable", true)) {
1281 port_s->enable = false;
1284 port_s->enable = true;
1287 /* STP over bonds is not supported. */
1288 if (!list_is_singleton(&port->ifaces)) {
1289 VLOG_ERR("port %s: cannot enable STP on bonds, disabling",
1291 port_s->enable = false;
1295 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1297 /* Internal ports shouldn't participate in spanning tree, so
1299 if (!strcmp(iface->type, "internal")) {
1300 VLOG_DBG("port %s: disable STP on internal ports", port->name);
1301 port_s->enable = false;
1305 /* STP on mirror output ports is not supported. */
1306 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1307 VLOG_DBG("port %s: disable STP on mirror ports", port->name);
1308 port_s->enable = false;
1312 config_str = smap_get(&port->cfg->other_config, "stp-port-num");
1314 unsigned long int port_num = strtoul(config_str, NULL, 0);
1315 int port_idx = port_num - 1;
1317 if (port_num < 1 || port_num > STP_MAX_PORTS) {
1318 VLOG_ERR("port %s: invalid stp-port-num", port->name);
1319 port_s->enable = false;
1323 if (bitmap_is_set(port_num_bitmap, port_idx)) {
1324 VLOG_ERR("port %s: duplicate stp-port-num %lu, disabling",
1325 port->name, port_num);
1326 port_s->enable = false;
1329 bitmap_set1(port_num_bitmap, port_idx);
1330 port_s->port_num = port_idx;
1332 if (*port_num_counter >= STP_MAX_PORTS) {
1333 VLOG_ERR("port %s: too many STP ports, disabling", port->name);
1334 port_s->enable = false;
1338 port_s->port_num = (*port_num_counter)++;
1341 config_str = smap_get(&port->cfg->other_config, "stp-path-cost");
1343 port_s->path_cost = strtoul(config_str, NULL, 10);
1345 enum netdev_features current;
1348 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1349 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1350 port_s->path_cost = stp_convert_speed_to_cost(mbps);
1353 config_str = smap_get(&port->cfg->other_config, "stp-port-priority");
1355 port_s->priority = strtoul(config_str, NULL, 0);
1357 port_s->priority = STP_DEFAULT_PORT_PRIORITY;
1362 port_configure_rstp(const struct ofproto *ofproto, struct port *port,
1363 struct ofproto_port_rstp_settings *port_s, int *port_num_counter)
1365 const char *config_str;
1366 struct iface *iface;
1368 if (!smap_get_bool(&port->cfg->other_config, "rstp-enable", true)) {
1369 port_s->enable = false;
1372 port_s->enable = true;
1375 /* RSTP over bonds is not supported. */
1376 if (!list_is_singleton(&port->ifaces)) {
1377 VLOG_ERR("port %s: cannot enable RSTP on bonds, disabling",
1379 port_s->enable = false;
1383 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
1385 /* Internal ports shouldn't participate in spanning tree, so
1387 if (!strcmp(iface->type, "internal")) {
1388 VLOG_DBG("port %s: disable RSTP on internal ports", port->name);
1389 port_s->enable = false;
1393 /* RSTP on mirror output ports is not supported. */
1394 if (ofproto_is_mirror_output_bundle(ofproto, port)) {
1395 VLOG_DBG("port %s: disable RSTP on mirror ports", port->name);
1396 port_s->enable = false;
1400 config_str = smap_get(&port->cfg->other_config, "rstp-port-num");
1402 unsigned long int port_num = strtoul(config_str, NULL, 0);
1403 if (port_num < 1 || port_num > RSTP_MAX_PORTS) {
1404 VLOG_ERR("port %s: invalid rstp-port-num", port->name);
1405 port_s->enable = false;
1408 port_s->port_num = port_num;
1410 if (*port_num_counter >= RSTP_MAX_PORTS) {
1411 VLOG_ERR("port %s: too many RSTP ports, disabling", port->name);
1412 port_s->enable = false;
1415 /* If rstp-port-num is not specified, use 0.
1416 * rstp_port_set_port_number() will look for the first free one. */
1417 port_s->port_num = 0;
1420 config_str = smap_get(&port->cfg->other_config, "rstp-path-cost");
1422 port_s->path_cost = strtoul(config_str, NULL, 10);
1424 enum netdev_features current;
1427 netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
1428 mbps = netdev_features_to_bps(current, 100 * 1000 * 1000) / 1000000;
1429 port_s->path_cost = rstp_convert_speed_to_cost(mbps);
1432 config_str = smap_get(&port->cfg->other_config, "rstp-port-priority");
1434 port_s->priority = strtoul(config_str, NULL, 0);
1436 port_s->priority = RSTP_DEFAULT_PORT_PRIORITY;
1439 config_str = smap_get(&port->cfg->other_config, "rstp-admin-p2p-mac");
1441 port_s->admin_p2p_mac_state = strtoul(config_str, NULL, 0);
1443 port_s->admin_p2p_mac_state = RSTP_ADMIN_P2P_MAC_FORCE_TRUE;
1446 port_s->admin_port_state = smap_get_bool(&port->cfg->other_config,
1447 "rstp-admin-port-state", true);
1449 port_s->admin_edge_port = smap_get_bool(&port->cfg->other_config,
1450 "rstp-port-admin-edge", false);
1451 port_s->auto_edge = smap_get_bool(&port->cfg->other_config,
1452 "rstp-port-auto-edge", true);
1453 port_s->mcheck = smap_get_bool(&port->cfg->other_config,
1454 "rstp-port-mcheck", false);
1457 /* Set spanning tree configuration on 'br'. */
1459 bridge_configure_stp(struct bridge *br, bool enable_stp)
1462 ofproto_set_stp(br->ofproto, NULL);
1464 struct ofproto_stp_settings br_s;
1465 const char *config_str;
1467 int port_num_counter;
1468 unsigned long *port_num_bitmap;
1470 config_str = smap_get(&br->cfg->other_config, "stp-system-id");
1472 uint8_t ea[ETH_ADDR_LEN];
1474 if (eth_addr_from_string(config_str, ea)) {
1475 br_s.system_id = eth_addr_to_uint64(ea);
1477 br_s.system_id = eth_addr_to_uint64(br->ea);
1478 VLOG_ERR("bridge %s: invalid stp-system-id, defaulting "
1479 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1482 br_s.system_id = eth_addr_to_uint64(br->ea);
1485 config_str = smap_get(&br->cfg->other_config, "stp-priority");
1487 br_s.priority = strtoul(config_str, NULL, 0);
1489 br_s.priority = STP_DEFAULT_BRIDGE_PRIORITY;
1492 config_str = smap_get(&br->cfg->other_config, "stp-hello-time");
1494 br_s.hello_time = strtoul(config_str, NULL, 10) * 1000;
1496 br_s.hello_time = STP_DEFAULT_HELLO_TIME;
1499 config_str = smap_get(&br->cfg->other_config, "stp-max-age");
1501 br_s.max_age = strtoul(config_str, NULL, 10) * 1000;
1503 br_s.max_age = STP_DEFAULT_MAX_AGE;
1506 config_str = smap_get(&br->cfg->other_config, "stp-forward-delay");
1508 br_s.fwd_delay = strtoul(config_str, NULL, 10) * 1000;
1510 br_s.fwd_delay = STP_DEFAULT_FWD_DELAY;
1513 /* Configure STP on the bridge. */
1514 if (ofproto_set_stp(br->ofproto, &br_s)) {
1515 VLOG_ERR("bridge %s: could not enable STP", br->name);
1519 /* Users must either set the port number with the "stp-port-num"
1520 * configuration on all ports or none. If manual configuration
1521 * is not done, then we allocate them sequentially. */
1522 port_num_counter = 0;
1523 port_num_bitmap = bitmap_allocate(STP_MAX_PORTS);
1524 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1525 struct ofproto_port_stp_settings port_s;
1526 struct iface *iface;
1528 port_configure_stp(br->ofproto, port, &port_s,
1529 &port_num_counter, port_num_bitmap);
1531 /* As bonds are not supported, just apply configuration to
1532 * all interfaces. */
1533 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1534 if (ofproto_port_set_stp(br->ofproto, iface->ofp_port,
1536 VLOG_ERR("port %s: could not enable STP", port->name);
1542 if (bitmap_scan(port_num_bitmap, 1, 0, STP_MAX_PORTS) != STP_MAX_PORTS
1543 && port_num_counter) {
1544 VLOG_ERR("bridge %s: must manually configure all STP port "
1545 "IDs or none, disabling", br->name);
1546 ofproto_set_stp(br->ofproto, NULL);
1548 bitmap_free(port_num_bitmap);
1553 bridge_configure_rstp(struct bridge *br, bool enable_rstp)
1556 ofproto_set_rstp(br->ofproto, NULL);
1558 struct ofproto_rstp_settings br_s;
1559 const char *config_str;
1561 int port_num_counter;
1563 config_str = smap_get(&br->cfg->other_config, "rstp-address");
1565 uint8_t ea[ETH_ADDR_LEN];
1567 if (eth_addr_from_string(config_str, ea)) {
1568 br_s.address = eth_addr_to_uint64(ea);
1571 br_s.address = eth_addr_to_uint64(br->ea);
1572 VLOG_ERR("bridge %s: invalid rstp-address, defaulting "
1573 "to "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(br->ea));
1577 br_s.address = eth_addr_to_uint64(br->ea);
1580 config_str = smap_get(&br->cfg->other_config, "rstp-priority");
1582 br_s.priority = strtoul(config_str, NULL, 0);
1584 br_s.priority = RSTP_DEFAULT_PRIORITY;
1587 config_str = smap_get(&br->cfg->other_config, "rstp-ageing-time");
1589 br_s.ageing_time = strtoul(config_str, NULL, 0);
1591 br_s.ageing_time = RSTP_DEFAULT_AGEING_TIME;
1594 config_str = smap_get(&br->cfg->other_config,
1595 "rstp-force-protocol-version");
1597 br_s.force_protocol_version = strtoul(config_str, NULL, 0);
1599 br_s.force_protocol_version = FPV_DEFAULT;
1602 config_str = smap_get(&br->cfg->other_config, "rstp-max-age");
1604 br_s.bridge_max_age = strtoul(config_str, NULL, 10);
1606 br_s.bridge_max_age = RSTP_DEFAULT_BRIDGE_MAX_AGE;
1609 config_str = smap_get(&br->cfg->other_config, "rstp-forward-delay");
1611 br_s.bridge_forward_delay = strtoul(config_str, NULL, 10);
1613 br_s.bridge_forward_delay = RSTP_DEFAULT_BRIDGE_FORWARD_DELAY;
1616 config_str = smap_get(&br->cfg->other_config,
1617 "rstp-transmit-hold-count");
1619 br_s.transmit_hold_count = strtoul(config_str, NULL, 10);
1621 br_s.transmit_hold_count = RSTP_DEFAULT_TRANSMIT_HOLD_COUNT;
1624 /* Configure RSTP on the bridge. */
1625 if (ofproto_set_rstp(br->ofproto, &br_s)) {
1626 VLOG_ERR("bridge %s: could not enable RSTP", br->name);
1630 port_num_counter = 0;
1631 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1632 struct ofproto_port_rstp_settings port_s;
1633 struct iface *iface;
1635 port_configure_rstp(br->ofproto, port, &port_s,
1638 /* As bonds are not supported, just apply configuration to
1639 * all interfaces. */
1640 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
1641 if (ofproto_port_set_rstp(br->ofproto, iface->ofp_port,
1643 VLOG_ERR("port %s: could not enable RSTP", port->name);
1652 bridge_configure_spanning_tree(struct bridge *br)
1654 bool enable_rstp = br->cfg->rstp_enable;
1655 bool enable_stp = br->cfg->stp_enable;
1657 if (enable_rstp && enable_stp) {
1658 VLOG_WARN("%s: RSTP and STP are mutually exclusive but both are "
1659 "configured; enabling RSTP", br->name);
1663 bridge_configure_stp(br, enable_stp);
1664 bridge_configure_rstp(br, enable_rstp);
1668 bridge_has_bond_fake_iface(const struct bridge *br, const char *name)
1670 const struct port *port = port_lookup(br, name);
1671 return port && port_is_bond_fake_iface(port);
1675 port_is_bond_fake_iface(const struct port *port)
1677 return port->cfg->bond_fake_iface && !list_is_short(&port->ifaces);
1681 add_del_bridges(const struct ovsrec_open_vswitch *cfg)
1683 struct bridge *br, *next;
1684 struct shash new_br;
1687 /* Collect new bridges' names and types. */
1688 shash_init(&new_br);
1689 for (i = 0; i < cfg->n_bridges; i++) {
1690 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1691 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1693 if (strchr(br_cfg->name, '/')) {
1694 /* Prevent remote ovsdb-server users from accessing arbitrary
1695 * directories, e.g. consider a bridge named "../../../etc/". */
1696 VLOG_WARN_RL(&rl, "ignoring bridge with invalid name \"%s\"",
1698 } else if (!shash_add_once(&new_br, br_cfg->name, br_cfg)) {
1699 VLOG_WARN_RL(&rl, "bridge %s specified twice", br_cfg->name);
1703 /* Get rid of deleted bridges or those whose types have changed.
1704 * Update 'cfg' of bridges that still exist. */
1705 HMAP_FOR_EACH_SAFE (br, next, node, &all_bridges) {
1706 br->cfg = shash_find_data(&new_br, br->name);
1707 if (!br->cfg || strcmp(br->type, ofproto_normalize_type(
1708 br->cfg->datapath_type))) {
1713 /* Add new bridges. */
1714 for (i = 0; i < cfg->n_bridges; i++) {
1715 const struct ovsrec_bridge *br_cfg = cfg->bridges[i];
1716 struct bridge *br = bridge_lookup(br_cfg->name);
1718 bridge_create(br_cfg);
1722 shash_destroy(&new_br);
1725 /* Configures 'netdev' based on the "options" column in 'iface_cfg'.
1726 * Returns 0 if successful, otherwise a positive errno value. */
1728 iface_set_netdev_config(const struct ovsrec_interface *iface_cfg,
1729 struct netdev *netdev, char **errp)
1731 return netdev_set_config(netdev, &iface_cfg->options, errp);
1734 /* Opens a network device for 'if_cfg' and configures it. Adds the network
1735 * device to br->ofproto and stores the OpenFlow port number in '*ofp_portp'.
1737 * If successful, returns 0 and stores the network device in '*netdevp'. On
1738 * failure, returns a positive errno value and stores NULL in '*netdevp'. */
1740 iface_do_create(const struct bridge *br,
1741 const struct ovsrec_interface *iface_cfg,
1742 const struct ovsrec_port *port_cfg,
1743 ofp_port_t *ofp_portp, struct netdev **netdevp,
1746 struct netdev *netdev = NULL;
1749 if (netdev_is_reserved_name(iface_cfg->name)) {
1750 VLOG_WARN("could not create interface %s, name is reserved",
1756 error = netdev_open(iface_cfg->name,
1757 iface_get_type(iface_cfg, br->cfg), &netdev);
1759 VLOG_WARN_BUF(errp, "could not open network device %s (%s)",
1760 iface_cfg->name, ovs_strerror(error));
1764 error = iface_set_netdev_config(iface_cfg, netdev, errp);
1769 *ofp_portp = iface_pick_ofport(iface_cfg);
1770 error = ofproto_port_add(br->ofproto, netdev, ofp_portp);
1775 VLOG_INFO("bridge %s: added interface %s on port %d",
1776 br->name, iface_cfg->name, *ofp_portp);
1778 if (port_cfg->vlan_mode && !strcmp(port_cfg->vlan_mode, "splinter")) {
1779 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
1787 netdev_close(netdev);
1791 /* Creates a new iface on 'br' based on 'if_cfg'. The new iface has OpenFlow
1792 * port number 'ofp_port'. If ofp_port is OFPP_NONE, an OpenFlow port is
1793 * automatically allocated for the iface. Takes ownership of and
1794 * deallocates 'if_cfg'.
1796 * Return true if an iface is successfully created, false otherwise. */
1798 iface_create(struct bridge *br, const struct ovsrec_interface *iface_cfg,
1799 const struct ovsrec_port *port_cfg)
1801 struct netdev *netdev;
1802 struct iface *iface;
1803 ofp_port_t ofp_port;
1808 /* Do the bits that can fail up front. */
1809 ovs_assert(!iface_lookup(br, iface_cfg->name));
1810 error = iface_do_create(br, iface_cfg, port_cfg, &ofp_port, &netdev, &errp);
1812 iface_clear_db_record(iface_cfg, errp);
1817 /* Get or create the port structure. */
1818 port = port_lookup(br, port_cfg->name);
1820 port = port_create(br, port_cfg);
1823 /* Create the iface structure. */
1824 iface = xzalloc(sizeof *iface);
1825 list_push_back(&port->ifaces, &iface->port_elem);
1826 hmap_insert(&br->iface_by_name, &iface->name_node,
1827 hash_string(iface_cfg->name, 0));
1829 iface->name = xstrdup(iface_cfg->name);
1830 iface->ofp_port = ofp_port;
1831 iface->netdev = netdev;
1832 iface->type = iface_get_type(iface_cfg, br->cfg);
1833 iface->cfg = iface_cfg;
1834 hmap_insert(&br->ifaces, &iface->ofp_port_node,
1835 hash_ofp_port(ofp_port));
1837 /* Populate initial status in database. */
1838 iface_refresh_stats(iface);
1839 iface_refresh_netdev_status(iface);
1841 /* Add bond fake iface if necessary. */
1842 if (port_is_bond_fake_iface(port)) {
1843 struct ofproto_port ofproto_port;
1845 if (ofproto_port_query_by_name(br->ofproto, port->name,
1847 struct netdev *netdev;
1850 error = netdev_open(port->name, "internal", &netdev);
1852 ofp_port_t fake_ofp_port = OFPP_NONE;
1853 ofproto_port_add(br->ofproto, netdev, &fake_ofp_port);
1854 netdev_close(netdev);
1856 VLOG_WARN("could not open network device %s (%s)",
1857 port->name, ovs_strerror(error));
1860 /* Already exists, nothing to do. */
1861 ofproto_port_destroy(&ofproto_port);
1868 /* Set forward BPDU option. */
1870 bridge_configure_forward_bpdu(struct bridge *br)
1872 ofproto_set_forward_bpdu(br->ofproto,
1873 smap_get_bool(&br->cfg->other_config,
1878 /* Set MAC learning table configuration for 'br'. */
1880 bridge_configure_mac_table(struct bridge *br)
1882 const char *idle_time_str;
1885 const char *mac_table_size_str;
1888 idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
1889 idle_time = (idle_time_str && atoi(idle_time_str)
1890 ? atoi(idle_time_str)
1891 : MAC_ENTRY_DEFAULT_IDLE_TIME);
1893 mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
1894 mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
1895 ? atoi(mac_table_size_str)
1898 ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
1901 /* Set multicast snooping table configuration for 'br'. */
1903 bridge_configure_mcast_snooping(struct bridge *br)
1905 if (!br->cfg->mcast_snooping_enable) {
1906 ofproto_set_mcast_snooping(br->ofproto, NULL);
1909 struct ofproto_mcast_snooping_settings br_s;
1910 const char *idle_time_str;
1911 const char *max_entries_str;
1913 idle_time_str = smap_get(&br->cfg->other_config,
1914 "mcast-snooping-aging-time");
1915 br_s.idle_time = (idle_time_str && atoi(idle_time_str)
1916 ? atoi(idle_time_str)
1917 : MCAST_ENTRY_DEFAULT_IDLE_TIME);
1919 max_entries_str = smap_get(&br->cfg->other_config,
1920 "mcast-snooping-table-size");
1921 br_s.max_entries = (max_entries_str && atoi(max_entries_str)
1922 ? atoi(max_entries_str)
1923 : MCAST_DEFAULT_MAX_ENTRIES);
1925 br_s.flood_unreg = !smap_get_bool(&br->cfg->other_config,
1926 "mcast-snooping-disable-flood-unregistered",
1929 /* Configure multicast snooping on the bridge */
1930 if (ofproto_set_mcast_snooping(br->ofproto, &br_s)) {
1931 VLOG_ERR("bridge %s: could not enable multicast snooping",
1936 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1937 struct ofproto_mcast_snooping_port_settings port_s;
1938 port_s.flood = smap_get_bool(&port->cfg->other_config,
1939 "mcast-snooping-flood", false);
1940 port_s.flood_reports = smap_get_bool(&port->cfg->other_config,
1941 "mcast-snooping-flood-reports", false);
1942 if (ofproto_port_set_mcast_snooping(br->ofproto, port, &port_s)) {
1943 VLOG_ERR("port %s: could not configure mcast snooping",
1951 find_local_hw_addr(const struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
1952 const struct port *fake_br, struct iface **hw_addr_iface)
1954 struct hmapx mirror_output_ports;
1956 bool found_addr = false;
1960 /* Mirror output ports don't participate in picking the local hardware
1961 * address. ofproto can't help us find out whether a given port is a
1962 * mirror output because we haven't configured mirrors yet, so we need to
1963 * accumulate them ourselves. */
1964 hmapx_init(&mirror_output_ports);
1965 for (i = 0; i < br->cfg->n_mirrors; i++) {
1966 struct ovsrec_mirror *m = br->cfg->mirrors[i];
1967 if (m->output_port) {
1968 hmapx_add(&mirror_output_ports, m->output_port);
1972 /* Otherwise choose the minimum non-local MAC address among all of the
1974 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
1975 uint8_t iface_ea[ETH_ADDR_LEN];
1976 struct iface *candidate;
1977 struct iface *iface;
1979 /* Mirror output ports don't participate. */
1980 if (hmapx_contains(&mirror_output_ports, port->cfg)) {
1984 /* Choose the MAC address to represent the port. */
1986 if (port->cfg->mac && eth_addr_from_string(port->cfg->mac, iface_ea)) {
1987 /* Find the interface with this Ethernet address (if any) so that
1988 * we can provide the correct devname to the caller. */
1989 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
1990 uint8_t candidate_ea[ETH_ADDR_LEN];
1991 if (!netdev_get_etheraddr(candidate->netdev, candidate_ea)
1992 && eth_addr_equals(iface_ea, candidate_ea)) {
1997 /* Choose the interface whose MAC address will represent the port.
1998 * The Linux kernel bonding code always chooses the MAC address of
1999 * the first slave added to a bond, and the Fedora networking
2000 * scripts always add slaves to a bond in alphabetical order, so
2001 * for compatibility we choose the interface with the name that is
2002 * first in alphabetical order. */
2003 LIST_FOR_EACH (candidate, port_elem, &port->ifaces) {
2004 if (!iface || strcmp(candidate->name, iface->name) < 0) {
2009 /* The local port doesn't count (since we're trying to choose its
2010 * MAC address anyway). */
2011 if (iface->ofp_port == OFPP_LOCAL) {
2015 /* For fake bridges we only choose from ports with the same tag */
2016 if (fake_br && fake_br->cfg && fake_br->cfg->tag) {
2017 if (!port->cfg->tag) {
2020 if (*port->cfg->tag != *fake_br->cfg->tag) {
2026 error = netdev_get_etheraddr(iface->netdev, iface_ea);
2032 /* Compare against our current choice. */
2033 if (!eth_addr_is_multicast(iface_ea) &&
2034 !eth_addr_is_local(iface_ea) &&
2035 !eth_addr_is_reserved(iface_ea) &&
2036 !eth_addr_is_zero(iface_ea) &&
2037 (!found_addr || eth_addr_compare_3way(iface_ea, ea) < 0))
2039 memcpy(ea, iface_ea, ETH_ADDR_LEN);
2040 *hw_addr_iface = iface;
2046 memcpy(ea, br->default_ea, ETH_ADDR_LEN);
2047 *hw_addr_iface = NULL;
2050 hmapx_destroy(&mirror_output_ports);
2054 bridge_pick_local_hw_addr(struct bridge *br, uint8_t ea[ETH_ADDR_LEN],
2055 struct iface **hw_addr_iface)
2058 *hw_addr_iface = NULL;
2060 /* Did the user request a particular MAC? */
2061 hwaddr = smap_get(&br->cfg->other_config, "hwaddr");
2062 if (hwaddr && eth_addr_from_string(hwaddr, ea)) {
2063 if (eth_addr_is_multicast(ea)) {
2064 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
2065 "address "ETH_ADDR_FMT, br->name, ETH_ADDR_ARGS(ea));
2066 } else if (eth_addr_is_zero(ea)) {
2067 VLOG_ERR("bridge %s: cannot set MAC address to zero", br->name);
2073 /* Find a local hw address */
2074 find_local_hw_addr(br, ea, NULL, hw_addr_iface);
2077 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
2078 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
2079 * an interface on 'br', then that interface must be passed in as
2080 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
2081 * 'hw_addr_iface' must be passed in as a null pointer. */
2083 bridge_pick_datapath_id(struct bridge *br,
2084 const uint8_t bridge_ea[ETH_ADDR_LEN],
2085 struct iface *hw_addr_iface)
2088 * The procedure for choosing a bridge MAC address will, in the most
2089 * ordinary case, also choose a unique MAC that we can use as a datapath
2090 * ID. In some special cases, though, multiple bridges will end up with
2091 * the same MAC address. This is OK for the bridges, but it will confuse
2092 * the OpenFlow controller, because each datapath needs a unique datapath
2095 * Datapath IDs must be unique. It is also very desirable that they be
2096 * stable from one run to the next, so that policy set on a datapath
2099 const char *datapath_id;
2102 datapath_id = smap_get(&br->cfg->other_config, "datapath-id");
2103 if (datapath_id && dpid_from_string(datapath_id, &dpid)) {
2107 if (!hw_addr_iface) {
2109 * A purely internal bridge, that is, one that has no non-virtual
2110 * network devices on it at all, is difficult because it has no
2111 * natural unique identifier at all.
2113 * When the host is a XenServer, we handle this case by hashing the
2114 * host's UUID with the name of the bridge. Names of bridges are
2115 * persistent across XenServer reboots, although they can be reused if
2116 * an internal network is destroyed and then a new one is later
2117 * created, so this is fairly effective.
2119 * When the host is not a XenServer, we punt by using a random MAC
2120 * address on each run.
2122 const char *host_uuid = xenserver_get_host_uuid();
2124 char *combined = xasprintf("%s,%s", host_uuid, br->name);
2125 dpid = dpid_from_hash(combined, strlen(combined));
2131 return eth_addr_to_uint64(bridge_ea);
2135 dpid_from_hash(const void *data, size_t n)
2137 uint8_t hash[SHA1_DIGEST_SIZE];
2139 BUILD_ASSERT_DECL(sizeof hash >= ETH_ADDR_LEN);
2140 sha1_bytes(data, n, hash);
2141 eth_addr_mark_random(hash);
2142 return eth_addr_to_uint64(hash);
2146 iface_refresh_netdev_status(struct iface *iface)
2150 enum netdev_features current;
2151 enum netdev_flags flags;
2152 const char *link_state;
2153 uint8_t mac[ETH_ADDR_LEN];
2154 int64_t bps, mtu_64, ifindex64, link_resets;
2157 if (iface_is_synthetic(iface)) {
2161 if (iface->change_seq == netdev_get_change_seq(iface->netdev)
2162 && !status_txn_try_again) {
2166 iface->change_seq = netdev_get_change_seq(iface->netdev);
2170 if (!netdev_get_status(iface->netdev, &smap)) {
2171 ovsrec_interface_set_status(iface->cfg, &smap);
2173 ovsrec_interface_set_status(iface->cfg, NULL);
2176 smap_destroy(&smap);
2178 error = netdev_get_flags(iface->netdev, &flags);
2180 const char *state = flags & NETDEV_UP ? "up" : "down";
2182 ovsrec_interface_set_admin_state(iface->cfg, state);
2184 ovsrec_interface_set_admin_state(iface->cfg, NULL);
2187 link_state = netdev_get_carrier(iface->netdev) ? "up" : "down";
2188 ovsrec_interface_set_link_state(iface->cfg, link_state);
2190 link_resets = netdev_get_carrier_resets(iface->netdev);
2191 ovsrec_interface_set_link_resets(iface->cfg, &link_resets, 1);
2193 error = netdev_get_features(iface->netdev, ¤t, NULL, NULL, NULL);
2194 bps = !error ? netdev_features_to_bps(current, 0) : 0;
2196 ovsrec_interface_set_duplex(iface->cfg,
2197 netdev_features_is_full_duplex(current)
2199 ovsrec_interface_set_link_speed(iface->cfg, &bps, 1);
2201 ovsrec_interface_set_duplex(iface->cfg, NULL);
2202 ovsrec_interface_set_link_speed(iface->cfg, NULL, 0);
2205 error = netdev_get_mtu(iface->netdev, &mtu);
2208 ovsrec_interface_set_mtu(iface->cfg, &mtu_64, 1);
2210 ovsrec_interface_set_mtu(iface->cfg, NULL, 0);
2213 error = netdev_get_etheraddr(iface->netdev, mac);
2215 char mac_string[32];
2217 sprintf(mac_string, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2218 ovsrec_interface_set_mac_in_use(iface->cfg, mac_string);
2220 ovsrec_interface_set_mac_in_use(iface->cfg, NULL);
2223 /* The netdev may return a negative number (such as -EOPNOTSUPP)
2224 * if there is no valid ifindex number. */
2225 ifindex64 = netdev_get_ifindex(iface->netdev);
2226 if (ifindex64 < 0) {
2229 ovsrec_interface_set_ifindex(iface->cfg, &ifindex64, 1);
2233 iface_refresh_ofproto_status(struct iface *iface)
2237 if (iface_is_synthetic(iface)) {
2241 current = ofproto_port_is_lacp_current(iface->port->bridge->ofproto,
2245 ovsrec_interface_set_lacp_current(iface->cfg, &bl, 1);
2247 ovsrec_interface_set_lacp_current(iface->cfg, NULL, 0);
2250 if (ofproto_port_cfm_status_changed(iface->port->bridge->ofproto,
2252 || status_txn_try_again) {
2253 iface_refresh_cfm_stats(iface);
2256 if (ofproto_port_bfd_status_changed(iface->port->bridge->ofproto,
2258 || status_txn_try_again) {
2262 ofproto_port_get_bfd_status(iface->port->bridge->ofproto,
2263 iface->ofp_port, &smap);
2264 ovsrec_interface_set_bfd_status(iface->cfg, &smap);
2265 smap_destroy(&smap);
2269 /* Writes 'iface''s CFM statistics to the database. 'iface' must not be
2272 iface_refresh_cfm_stats(struct iface *iface)
2274 const struct ovsrec_interface *cfg = iface->cfg;
2275 struct cfm_status status;
2278 error = ofproto_port_get_cfm_status(iface->port->bridge->ofproto,
2279 iface->ofp_port, &status);
2281 ovsrec_interface_set_cfm_fault(cfg, NULL, 0);
2282 ovsrec_interface_set_cfm_fault_status(cfg, NULL, 0);
2283 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2284 ovsrec_interface_set_cfm_flap_count(cfg, NULL, 0);
2285 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2286 ovsrec_interface_set_cfm_remote_mpids(cfg, NULL, 0);
2288 const char *reasons[CFM_FAULT_N_REASONS];
2289 int64_t cfm_health = status.health;
2290 int64_t cfm_flap_count = status.flap_count;
2291 bool faulted = status.faults != 0;
2294 ovsrec_interface_set_cfm_fault(cfg, &faulted, 1);
2297 for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
2298 int reason = 1 << i;
2299 if (status.faults & reason) {
2300 reasons[j++] = cfm_fault_reason_to_str(reason);
2303 ovsrec_interface_set_cfm_fault_status(cfg, reasons, j);
2305 ovsrec_interface_set_cfm_flap_count(cfg, &cfm_flap_count, 1);
2307 if (status.remote_opstate >= 0) {
2308 const char *remote_opstate = status.remote_opstate ? "up" : "down";
2309 ovsrec_interface_set_cfm_remote_opstate(cfg, remote_opstate);
2311 ovsrec_interface_set_cfm_remote_opstate(cfg, NULL);
2314 ovsrec_interface_set_cfm_remote_mpids(cfg,
2315 (const int64_t *)status.rmps,
2317 if (cfm_health >= 0) {
2318 ovsrec_interface_set_cfm_health(cfg, &cfm_health, 1);
2320 ovsrec_interface_set_cfm_health(cfg, NULL, 0);
2328 iface_refresh_stats(struct iface *iface)
2330 #define IFACE_STATS \
2331 IFACE_STAT(rx_packets, "rx_packets") \
2332 IFACE_STAT(tx_packets, "tx_packets") \
2333 IFACE_STAT(rx_bytes, "rx_bytes") \
2334 IFACE_STAT(tx_bytes, "tx_bytes") \
2335 IFACE_STAT(rx_dropped, "rx_dropped") \
2336 IFACE_STAT(tx_dropped, "tx_dropped") \
2337 IFACE_STAT(rx_errors, "rx_errors") \
2338 IFACE_STAT(tx_errors, "tx_errors") \
2339 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
2340 IFACE_STAT(rx_over_errors, "rx_over_err") \
2341 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
2342 IFACE_STAT(collisions, "collisions")
2344 #define IFACE_STAT(MEMBER, NAME) + 1
2345 enum { N_IFACE_STATS = IFACE_STATS };
2347 int64_t values[N_IFACE_STATS];
2348 const char *keys[N_IFACE_STATS];
2351 struct netdev_stats stats;
2353 if (iface_is_synthetic(iface)) {
2357 /* Intentionally ignore return value, since errors will set 'stats' to
2358 * all-1s, and we will deal with that correctly below. */
2359 netdev_get_stats(iface->netdev, &stats);
2361 /* Copy statistics into keys[] and values[]. */
2363 #define IFACE_STAT(MEMBER, NAME) \
2364 if (stats.MEMBER != UINT64_MAX) { \
2366 values[n] = stats.MEMBER; \
2371 ovs_assert(n <= N_IFACE_STATS);
2373 ovsrec_interface_set_statistics(iface->cfg, keys, values, n);
2378 br_refresh_datapath_info(struct bridge *br)
2380 const char *version;
2382 version = (br->ofproto && br->ofproto->ofproto_class->get_datapath_version
2383 ? br->ofproto->ofproto_class->get_datapath_version(br->ofproto)
2386 ovsrec_bridge_set_datapath_version(br->cfg,
2387 version ? version : "<unknown>");
2391 br_refresh_stp_status(struct bridge *br)
2393 struct smap smap = SMAP_INITIALIZER(&smap);
2394 struct ofproto *ofproto = br->ofproto;
2395 struct ofproto_stp_status status;
2397 if (ofproto_get_stp_status(ofproto, &status)) {
2401 if (!status.enabled) {
2402 ovsrec_bridge_set_status(br->cfg, NULL);
2406 smap_add_format(&smap, "stp_bridge_id", STP_ID_FMT,
2407 STP_ID_ARGS(status.bridge_id));
2408 smap_add_format(&smap, "stp_designated_root", STP_ID_FMT,
2409 STP_ID_ARGS(status.designated_root));
2410 smap_add_format(&smap, "stp_root_path_cost", "%d", status.root_path_cost);
2412 ovsrec_bridge_set_status(br->cfg, &smap);
2413 smap_destroy(&smap);
2417 port_refresh_stp_status(struct port *port)
2419 struct ofproto *ofproto = port->bridge->ofproto;
2420 struct iface *iface;
2421 struct ofproto_port_stp_status status;
2424 if (port_is_synthetic(port)) {
2428 /* STP doesn't currently support bonds. */
2429 if (!list_is_singleton(&port->ifaces)) {
2430 ovsrec_port_set_status(port->cfg, NULL);
2434 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2435 if (ofproto_port_get_stp_status(ofproto, iface->ofp_port, &status)) {
2439 if (!status.enabled) {
2440 ovsrec_port_set_status(port->cfg, NULL);
2444 /* Set Status column. */
2446 smap_add_format(&smap, "stp_port_id", STP_PORT_ID_FMT, status.port_id);
2447 smap_add(&smap, "stp_state", stp_state_name(status.state));
2448 smap_add_format(&smap, "stp_sec_in_state", "%u", status.sec_in_state);
2449 smap_add(&smap, "stp_role", stp_role_name(status.role));
2450 ovsrec_port_set_status(port->cfg, &smap);
2451 smap_destroy(&smap);
2455 port_refresh_stp_stats(struct port *port)
2457 struct ofproto *ofproto = port->bridge->ofproto;
2458 struct iface *iface;
2459 struct ofproto_port_stp_stats stats;
2460 const char *keys[3];
2461 int64_t int_values[3];
2463 if (port_is_synthetic(port)) {
2467 /* STP doesn't currently support bonds. */
2468 if (!list_is_singleton(&port->ifaces)) {
2472 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2473 if (ofproto_port_get_stp_stats(ofproto, iface->ofp_port, &stats)) {
2477 if (!stats.enabled) {
2478 ovsrec_port_set_statistics(port->cfg, NULL, NULL, 0);
2482 /* Set Statistics column. */
2483 keys[0] = "stp_tx_count";
2484 int_values[0] = stats.tx_count;
2485 keys[1] = "stp_rx_count";
2486 int_values[1] = stats.rx_count;
2487 keys[2] = "stp_error_count";
2488 int_values[2] = stats.error_count;
2490 ovsrec_port_set_statistics(port->cfg, keys, int_values,
2491 ARRAY_SIZE(int_values));
2495 br_refresh_rstp_status(struct bridge *br)
2497 struct smap smap = SMAP_INITIALIZER(&smap);
2498 struct ofproto *ofproto = br->ofproto;
2499 struct ofproto_rstp_status status;
2501 if (ofproto_get_rstp_status(ofproto, &status)) {
2504 if (!status.enabled) {
2505 ovsrec_bridge_set_rstp_status(br->cfg, NULL);
2508 smap_add_format(&smap, "rstp_bridge_id", RSTP_ID_FMT,
2509 RSTP_ID_ARGS(status.bridge_id));
2510 smap_add_format(&smap, "rstp_root_path_cost", "%"PRIu32,
2511 status.root_path_cost);
2512 smap_add_format(&smap, "rstp_root_id", RSTP_ID_FMT,
2513 RSTP_ID_ARGS(status.root_id));
2514 smap_add_format(&smap, "rstp_designated_id", RSTP_ID_FMT,
2515 RSTP_ID_ARGS(status.designated_id));
2516 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2517 status.designated_port_id);
2518 smap_add_format(&smap, "rstp_bridge_port_id", RSTP_PORT_ID_FMT,
2519 status.bridge_port_id);
2520 ovsrec_bridge_set_rstp_status(br->cfg, &smap);
2521 smap_destroy(&smap);
2525 port_refresh_rstp_status(struct port *port)
2527 struct ofproto *ofproto = port->bridge->ofproto;
2528 struct iface *iface;
2529 struct ofproto_port_rstp_status status;
2530 const char *keys[4];
2531 int64_t int_values[4];
2534 if (port_is_synthetic(port)) {
2538 /* RSTP doesn't currently support bonds. */
2539 if (!list_is_singleton(&port->ifaces)) {
2540 ovsrec_port_set_rstp_status(port->cfg, NULL);
2544 iface = CONTAINER_OF(list_front(&port->ifaces), struct iface, port_elem);
2545 if (ofproto_port_get_rstp_status(ofproto, iface->ofp_port, &status)) {
2549 if (!status.enabled) {
2550 ovsrec_port_set_rstp_status(port->cfg, NULL);
2551 ovsrec_port_set_rstp_statistics(port->cfg, NULL, NULL, 0);
2554 /* Set Status column. */
2557 smap_add_format(&smap, "rstp_port_id", RSTP_PORT_ID_FMT,
2559 smap_add_format(&smap, "rstp_port_role", "%s",
2560 rstp_port_role_name(status.role));
2561 smap_add_format(&smap, "rstp_port_state", "%s",
2562 rstp_state_name(status.state));
2563 smap_add_format(&smap, "rstp_designated_bridge_id", RSTP_ID_FMT,
2564 RSTP_ID_ARGS(status.designated_bridge_id));
2565 smap_add_format(&smap, "rstp_designated_port_id", RSTP_PORT_ID_FMT,
2566 status.designated_port_id);
2567 smap_add_format(&smap, "rstp_designated_path_cost", "%"PRIu32,
2568 status.designated_path_cost);
2570 ovsrec_port_set_rstp_status(port->cfg, &smap);
2571 smap_destroy(&smap);
2573 /* Set Statistics column. */
2574 keys[0] = "rstp_tx_count";
2575 int_values[0] = status.tx_count;
2576 keys[1] = "rstp_rx_count";
2577 int_values[1] = status.rx_count;
2578 keys[2] = "rstp_uptime";
2579 int_values[2] = status.uptime;
2580 keys[3] = "rstp_error_count";
2581 int_values[3] = status.error_count;
2582 ovsrec_port_set_rstp_statistics(port->cfg, keys, int_values,
2583 ARRAY_SIZE(int_values));
2587 port_refresh_bond_status(struct port *port, bool force_update)
2589 uint8_t mac[ETH_ADDR_LEN];
2591 /* Return if port is not a bond */
2592 if (list_is_singleton(&port->ifaces)) {
2596 if (bond_get_changed_active_slave(port->name, mac, force_update)) {
2600 ds_put_format(&mac_s, ETH_ADDR_FMT, ETH_ADDR_ARGS(mac));
2601 ovsrec_port_set_bond_active_slave(port->cfg, ds_cstr(&mac_s));
2607 enable_system_stats(const struct ovsrec_open_vswitch *cfg)
2609 return smap_get_bool(&cfg->other_config, "enable-statistics", false);
2613 reconfigure_system_stats(const struct ovsrec_open_vswitch *cfg)
2615 bool enable = enable_system_stats(cfg);
2617 system_stats_enable(enable);
2619 ovsrec_open_vswitch_set_statistics(cfg, NULL);
2624 run_system_stats(void)
2626 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2629 stats = system_stats_run();
2631 struct ovsdb_idl_txn *txn;
2632 struct ovsdb_datum datum;
2634 txn = ovsdb_idl_txn_create(idl);
2635 ovsdb_datum_from_smap(&datum, stats);
2636 ovsdb_idl_txn_write(&cfg->header_, &ovsrec_open_vswitch_col_statistics,
2638 ovsdb_idl_txn_commit(txn);
2639 ovsdb_idl_txn_destroy(txn);
2646 ofp12_controller_role_to_str(enum ofp12_controller_role role)
2649 case OFPCR12_ROLE_EQUAL:
2651 case OFPCR12_ROLE_MASTER:
2653 case OFPCR12_ROLE_SLAVE:
2655 case OFPCR12_ROLE_NOCHANGE:
2657 return "*** INVALID ROLE ***";
2662 refresh_controller_status(void)
2666 const struct ovsrec_controller *cfg;
2670 /* Accumulate status for controllers on all bridges. */
2671 HMAP_FOR_EACH (br, node, &all_bridges) {
2672 ofproto_get_ofproto_controller_info(br->ofproto, &info);
2675 /* Update each controller in the database with current status. */
2676 OVSREC_CONTROLLER_FOR_EACH(cfg, idl) {
2677 struct ofproto_controller_info *cinfo =
2678 shash_find_data(&info, cfg->target);
2681 ovsrec_controller_set_is_connected(cfg, cinfo->is_connected);
2682 ovsrec_controller_set_role(cfg, ofp12_controller_role_to_str(
2684 ovsrec_controller_set_status(cfg, &cinfo->pairs);
2686 ovsrec_controller_set_is_connected(cfg, false);
2687 ovsrec_controller_set_role(cfg, NULL);
2688 ovsrec_controller_set_status(cfg, NULL);
2692 ofproto_free_ofproto_controller_info(&info);
2695 /* Update interface and mirror statistics if necessary. */
2697 run_stats_update(void)
2699 const struct ovsrec_open_vswitch *cfg = ovsrec_open_vswitch_first(idl);
2706 /* Statistics update interval should always be greater than or equal to
2708 stats_interval = MAX(smap_get_int(&cfg->other_config,
2709 "stats-update-interval",
2711 if (stats_timer_interval != stats_interval) {
2712 stats_timer_interval = stats_interval;
2713 stats_timer = LLONG_MIN;
2716 if (time_msec() >= stats_timer) {
2717 enum ovsdb_idl_txn_status status;
2719 /* Rate limit the update. Do not start a new update if the
2720 * previous one is not done. */
2724 stats_txn = ovsdb_idl_txn_create(idl);
2725 HMAP_FOR_EACH (br, node, &all_bridges) {
2729 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2730 struct iface *iface;
2732 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2733 iface_refresh_stats(iface);
2735 port_refresh_stp_stats(port);
2737 HMAP_FOR_EACH (m, hmap_node, &br->mirrors) {
2738 mirror_refresh_stats(m);
2741 refresh_controller_status();
2744 status = ovsdb_idl_txn_commit(stats_txn);
2745 if (status != TXN_INCOMPLETE) {
2746 stats_timer = time_msec() + stats_timer_interval;
2747 ovsdb_idl_txn_destroy(stats_txn);
2754 stats_update_wait(void)
2756 /* If the 'stats_txn' is non-null (transaction incomplete), waits for the
2757 * transaction to complete. Otherwise, waits for the 'stats_timer'. */
2759 ovsdb_idl_txn_wait(stats_txn);
2761 poll_timer_wait_until(stats_timer);
2765 /* Update bridge/port/interface status if necessary. */
2767 run_status_update(void)
2772 /* Rate limit the update. Do not start a new update if the
2773 * previous one is not done. */
2774 seq = seq_read(connectivity_seq_get());
2775 if (seq != connectivity_seqno || status_txn_try_again) {
2778 connectivity_seqno = seq;
2779 status_txn = ovsdb_idl_txn_create(idl);
2780 HMAP_FOR_EACH (br, node, &all_bridges) {
2783 br_refresh_stp_status(br);
2784 br_refresh_rstp_status(br);
2785 br_refresh_datapath_info(br);
2786 HMAP_FOR_EACH (port, hmap_node, &br->ports) {
2787 struct iface *iface;
2789 port_refresh_stp_status(port);
2790 port_refresh_rstp_status(port);
2791 port_refresh_bond_status(port, status_txn_try_again);
2792 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
2793 iface_refresh_netdev_status(iface);
2794 iface_refresh_ofproto_status(iface);
2801 /* Commit the transaction and get the status. If the transaction finishes,
2802 * then destroy the transaction. Otherwise, keep it so that we can check
2803 * progress the next time that this function is called. */
2805 enum ovsdb_idl_txn_status status;
2807 status = ovsdb_idl_txn_commit(status_txn);
2808 if (status != TXN_INCOMPLETE) {
2809 ovsdb_idl_txn_destroy(status_txn);
2812 /* Sets the 'status_txn_try_again' if the transaction fails. */
2813 if (status == TXN_SUCCESS || status == TXN_UNCHANGED) {
2814 status_txn_try_again = false;
2816 status_txn_try_again = true;
2821 /* Refresh AA port status if necessary. */
2822 if (time_msec() >= aa_refresh_timer) {
2825 HMAP_FOR_EACH (br, node, &all_bridges) {
2826 if (bridge_aa_need_refresh(br)) {
2827 struct ovsdb_idl_txn *txn;
2829 txn = ovsdb_idl_txn_create(idl);
2830 bridge_aa_refresh_queued(br);
2831 ovsdb_idl_txn_commit(txn);
2832 ovsdb_idl_txn_destroy(txn);
2836 aa_refresh_timer = time_msec() + AA_REFRESH_INTERVAL;
2841 status_update_wait(void)
2843 /* If the 'status_txn' is non-null (transaction incomplete), waits for the
2844 * transaction to complete. If the status update to database needs to be
2845 * run again (transaction fails), registers a timeout in
2846 * 'STATUS_CHECK_AGAIN_MSEC'. Otherwise, waits on the global connectivity
2847 * sequence number. */
2849 ovsdb_idl_txn_wait(status_txn);
2850 } else if (status_txn_try_again) {
2851 poll_timer_wait_until(time_msec() + STATUS_CHECK_AGAIN_MSEC);
2853 seq_wait(connectivity_seq_get(), connectivity_seqno);
2864 /* Let each datapath type do the work that it needs to do. */
2866 ofproto_enumerate_types(&types);
2867 SSET_FOR_EACH (type, &types) {
2868 ofproto_type_run(type);
2870 sset_destroy(&types);
2872 /* Let each bridge do the work that it needs to do. */
2873 HMAP_FOR_EACH (br, node, &all_bridges) {
2874 ofproto_run(br->ofproto);
2881 static struct ovsrec_open_vswitch null_cfg;
2882 const struct ovsrec_open_vswitch *cfg;
2884 bool vlan_splinters_changed;
2886 ovsrec_open_vswitch_init(&null_cfg);
2890 if (ovsdb_idl_is_lock_contended(idl)) {
2891 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2892 struct bridge *br, *next_br;
2894 VLOG_ERR_RL(&rl, "another ovs-vswitchd process is running, "
2895 "disabling this process (pid %ld) until it goes away",
2896 (long int) getpid());
2898 HMAP_FOR_EACH_SAFE (br, next_br, node, &all_bridges) {
2901 /* Since we will not be running system_stats_run() in this process
2902 * with the current situation of multiple ovs-vswitchd daemons,
2903 * disable system stats collection. */
2904 system_stats_enable(false);
2906 } else if (!ovsdb_idl_has_lock(idl)
2907 || !ovsdb_idl_has_ever_connected(idl)) {
2908 /* Returns if not holding the lock or not done retrieving db
2912 cfg = ovsrec_open_vswitch_first(idl);
2914 /* Initialize the ofproto library. This only needs to run once, but
2915 * it must be done after the configuration is set. If the
2916 * initialization has already occurred, bridge_init_ofproto()
2917 * returns immediately. */
2918 bridge_init_ofproto(cfg);
2920 /* Once the value of flow-restore-wait is false, we no longer should
2921 * check its value from the database. */
2922 if (cfg && ofproto_get_flow_restore_wait()) {
2923 ofproto_set_flow_restore_wait(smap_get_bool(&cfg->other_config,
2924 "flow-restore-wait", false));
2929 /* Re-configure SSL. We do this on every trip through the main loop,
2930 * instead of just when the database changes, because the contents of the
2931 * key and certificate files can change without the database changing.
2933 * We do this before bridge_reconfigure() because that function might
2934 * initiate SSL connections and thus requires SSL to be configured. */
2935 if (cfg && cfg->ssl) {
2936 const struct ovsrec_ssl *ssl = cfg->ssl;
2938 stream_ssl_set_key_and_cert(ssl->private_key, ssl->certificate);
2939 stream_ssl_set_ca_cert_file(ssl->ca_cert, ssl->bootstrap_ca_cert);
2942 /* If VLAN splinters are in use, then we need to reconfigure if VLAN
2943 * usage has changed. */
2944 vlan_splinters_changed = false;
2945 if (vlan_splinters_enabled_anywhere) {
2948 HMAP_FOR_EACH (br, node, &all_bridges) {
2949 if (ofproto_has_vlan_usage_changed(br->ofproto)) {
2950 vlan_splinters_changed = true;
2956 if (ovsdb_idl_get_seqno(idl) != idl_seqno || vlan_splinters_changed) {
2957 struct ovsdb_idl_txn *txn;
2959 idl_seqno = ovsdb_idl_get_seqno(idl);
2960 txn = ovsdb_idl_txn_create(idl);
2961 bridge_reconfigure(cfg ? cfg : &null_cfg);
2964 ovsrec_open_vswitch_set_cur_cfg(cfg, cfg->next_cfg);
2965 discover_types(cfg);
2968 /* If we are completing our initial configuration for this run
2969 * of ovs-vswitchd, then keep the transaction around to monitor
2970 * it for completion. */
2971 if (initial_config_done) {
2972 /* Always sets the 'status_txn_try_again' to check again,
2973 * in case that this transaction fails. */
2974 status_txn_try_again = true;
2975 ovsdb_idl_txn_commit(txn);
2976 ovsdb_idl_txn_destroy(txn);
2978 initial_config_done = true;
2979 daemonize_txn = txn;
2983 if (daemonize_txn) {
2984 enum ovsdb_idl_txn_status status = ovsdb_idl_txn_commit(daemonize_txn);
2985 if (status != TXN_INCOMPLETE) {
2986 ovsdb_idl_txn_destroy(daemonize_txn);
2987 daemonize_txn = NULL;
2989 /* ovs-vswitchd has completed initialization, so allow the
2990 * process that forked us to exit successfully. */
2991 daemonize_complete();
2993 vlog_enable_async();
2995 VLOG_INFO_ONCE("%s (Open vSwitch) %s", program_name, VERSION);
3000 run_status_update();
3010 ovsdb_idl_wait(idl);
3011 if (daemonize_txn) {
3012 ovsdb_idl_txn_wait(daemonize_txn);
3016 ofproto_enumerate_types(&types);
3017 SSET_FOR_EACH (type, &types) {
3018 ofproto_type_wait(type);
3020 sset_destroy(&types);
3022 if (!hmap_is_empty(&all_bridges)) {
3025 HMAP_FOR_EACH (br, node, &all_bridges) {
3026 ofproto_wait(br->ofproto);
3028 stats_update_wait();
3029 status_update_wait();
3032 system_stats_wait();
3035 /* Adds some memory usage statistics for bridges into 'usage', for use with
3036 * memory_report(). */
3038 bridge_get_memory_usage(struct simap *usage)
3045 ofproto_enumerate_types(&types);
3046 SSET_FOR_EACH (type, &types) {
3047 ofproto_type_get_memory_usage(type, usage);
3049 sset_destroy(&types);
3051 HMAP_FOR_EACH (br, node, &all_bridges) {
3052 ofproto_get_memory_usage(br->ofproto, usage);
3056 /* QoS unixctl user interface functions. */
3058 struct qos_unixctl_show_cbdata {
3060 struct iface *iface;
3064 qos_unixctl_show_queue(unsigned int queue_id,
3065 const struct smap *details,
3066 struct iface *iface,
3069 struct netdev_queue_stats stats;
3070 struct smap_node *node;
3073 ds_put_cstr(ds, "\n");
3075 ds_put_format(ds, "Queue %u:\n", queue_id);
3077 ds_put_cstr(ds, "Default:\n");
3080 SMAP_FOR_EACH (node, details) {
3081 ds_put_format(ds, "\t%s: %s\n", node->key, node->value);
3084 error = netdev_get_queue_stats(iface->netdev, queue_id, &stats);
3086 if (stats.tx_packets != UINT64_MAX) {
3087 ds_put_format(ds, "\ttx_packets: %"PRIu64"\n", stats.tx_packets);
3090 if (stats.tx_bytes != UINT64_MAX) {
3091 ds_put_format(ds, "\ttx_bytes: %"PRIu64"\n", stats.tx_bytes);
3094 if (stats.tx_errors != UINT64_MAX) {
3095 ds_put_format(ds, "\ttx_errors: %"PRIu64"\n", stats.tx_errors);
3098 ds_put_format(ds, "\tFailed to get statistics for queue %u: %s",
3099 queue_id, ovs_strerror(error));
3104 qos_unixctl_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
3105 const char *argv[], void *aux OVS_UNUSED)
3107 struct ds ds = DS_EMPTY_INITIALIZER;
3108 struct smap smap = SMAP_INITIALIZER(&smap);
3109 struct iface *iface;
3111 struct smap_node *node;
3113 iface = iface_find(argv[1]);
3115 unixctl_command_reply_error(conn, "no such interface");
3119 netdev_get_qos(iface->netdev, &type, &smap);
3121 if (*type != '\0') {
3122 struct netdev_queue_dump dump;
3123 struct smap details;
3124 unsigned int queue_id;
3126 ds_put_format(&ds, "QoS: %s %s\n", iface->name, type);
3128 SMAP_FOR_EACH (node, &smap) {
3129 ds_put_format(&ds, "%s: %s\n", node->key, node->value);
3132 smap_init(&details);
3133 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
3134 qos_unixctl_show_queue(queue_id, &details, iface, &ds);
3136 smap_destroy(&details);
3138 unixctl_command_reply(conn, ds_cstr(&ds));
3140 ds_put_format(&ds, "QoS not configured on %s\n", iface->name);
3141 unixctl_command_reply_error(conn, ds_cstr(&ds));
3144 smap_destroy(&smap);
3148 /* Bridge reconfiguration functions. */
3150 bridge_create(const struct ovsrec_bridge *br_cfg)
3154 ovs_assert(!bridge_lookup(br_cfg->name));
3155 br = xzalloc(sizeof *br);
3157 br->name = xstrdup(br_cfg->name);
3158 br->type = xstrdup(ofproto_normalize_type(br_cfg->datapath_type));
3161 /* Derive the default Ethernet address from the bridge's UUID. This should
3162 * be unique and it will be stable between ovs-vswitchd runs. */
3163 memcpy(br->default_ea, &br_cfg->header_.uuid, ETH_ADDR_LEN);
3164 eth_addr_mark_random(br->default_ea);
3166 hmap_init(&br->ports);
3167 hmap_init(&br->ifaces);
3168 hmap_init(&br->iface_by_name);
3169 hmap_init(&br->mirrors);
3171 hmap_init(&br->mappings);
3172 hmap_insert(&all_bridges, &br->node, hash_string(br->name, 0));
3176 bridge_destroy(struct bridge *br)
3179 struct mirror *mirror, *next_mirror;
3180 struct port *port, *next_port;
3182 HMAP_FOR_EACH_SAFE (port, next_port, hmap_node, &br->ports) {
3185 HMAP_FOR_EACH_SAFE (mirror, next_mirror, hmap_node, &br->mirrors) {
3186 mirror_destroy(mirror);
3189 hmap_remove(&all_bridges, &br->node);
3190 ofproto_destroy(br->ofproto);
3191 hmap_destroy(&br->ifaces);
3192 hmap_destroy(&br->ports);
3193 hmap_destroy(&br->iface_by_name);
3194 hmap_destroy(&br->mirrors);
3195 hmap_destroy(&br->mappings);
3202 static struct bridge *
3203 bridge_lookup(const char *name)
3207 HMAP_FOR_EACH_WITH_HASH (br, node, hash_string(name, 0), &all_bridges) {
3208 if (!strcmp(br->name, name)) {
3215 /* Handle requests for a listing of all flows known by the OpenFlow
3216 * stack, including those normally hidden. */
3218 bridge_unixctl_dump_flows(struct unixctl_conn *conn, int argc OVS_UNUSED,
3219 const char *argv[], void *aux OVS_UNUSED)
3224 br = bridge_lookup(argv[1]);
3226 unixctl_command_reply_error(conn, "Unknown bridge");
3231 ofproto_get_all_flows(br->ofproto, &results);
3233 unixctl_command_reply(conn, ds_cstr(&results));
3234 ds_destroy(&results);
3237 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
3238 * connections and reconnect. If BRIDGE is not specified, then all bridges
3239 * drop their controller connections and reconnect. */
3241 bridge_unixctl_reconnect(struct unixctl_conn *conn, int argc,
3242 const char *argv[], void *aux OVS_UNUSED)
3246 br = bridge_lookup(argv[1]);
3248 unixctl_command_reply_error(conn, "Unknown bridge");
3251 ofproto_reconnect_controllers(br->ofproto);
3253 HMAP_FOR_EACH (br, node, &all_bridges) {
3254 ofproto_reconnect_controllers(br->ofproto);
3257 unixctl_command_reply(conn, NULL);
3261 bridge_get_controllers(const struct bridge *br,
3262 struct ovsrec_controller ***controllersp)
3264 struct ovsrec_controller **controllers;
3265 size_t n_controllers;
3267 controllers = br->cfg->controller;
3268 n_controllers = br->cfg->n_controller;
3270 if (n_controllers == 1 && !strcmp(controllers[0]->target, "none")) {
3276 *controllersp = controllers;
3278 return n_controllers;
3282 bridge_collect_wanted_ports(struct bridge *br,
3283 const unsigned long int *splinter_vlans,
3284 struct shash *wanted_ports)
3288 shash_init(wanted_ports);
3290 for (i = 0; i < br->cfg->n_ports; i++) {
3291 const char *name = br->cfg->ports[i]->name;
3292 if (!shash_add_once(wanted_ports, name, br->cfg->ports[i])) {
3293 VLOG_WARN("bridge %s: %s specified twice as bridge port",
3297 if (bridge_get_controllers(br, NULL)
3298 && !shash_find(wanted_ports, br->name)) {
3299 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
3300 br->name, br->name);
3302 ovsrec_interface_init(&br->synth_local_iface);
3303 ovsrec_port_init(&br->synth_local_port);
3305 br->synth_local_port.interfaces = &br->synth_local_ifacep;
3306 br->synth_local_port.n_interfaces = 1;
3307 br->synth_local_port.name = br->name;
3309 br->synth_local_iface.name = br->name;
3310 br->synth_local_iface.type = "internal";
3312 br->synth_local_ifacep = &br->synth_local_iface;
3314 shash_add(wanted_ports, br->name, &br->synth_local_port);
3317 if (splinter_vlans) {
3318 add_vlan_splinter_ports(br, splinter_vlans, wanted_ports);
3322 /* Deletes "struct port"s and "struct iface"s under 'br' which aren't
3323 * consistent with 'br->cfg'. Updates 'br->if_cfg_queue' with interfaces which
3324 * 'br' needs to complete its configuration. */
3326 bridge_del_ports(struct bridge *br, const struct shash *wanted_ports)
3328 struct shash_node *port_node;
3329 struct port *port, *next;
3331 /* Get rid of deleted ports.
3332 * Get rid of deleted interfaces on ports that still exist. */
3333 HMAP_FOR_EACH_SAFE (port, next, hmap_node, &br->ports) {
3334 port->cfg = shash_find_data(wanted_ports, port->name);
3338 port_del_ifaces(port);
3342 /* Update iface->cfg and iface->type in interfaces that still exist. */
3343 SHASH_FOR_EACH (port_node, wanted_ports) {
3344 const struct ovsrec_port *port = port_node->data;
3347 for (i = 0; i < port->n_interfaces; i++) {
3348 const struct ovsrec_interface *cfg = port->interfaces[i];
3349 struct iface *iface = iface_lookup(br, cfg->name);
3350 const char *type = iface_get_type(cfg, br->cfg);
3355 } else if (!strcmp(type, "null")) {
3356 VLOG_WARN_ONCE("%s: The null interface type is deprecated and"
3357 " may be removed in February 2013. Please email"
3358 " dev@openvswitch.org with concerns.",
3361 /* We will add new interfaces later. */
3367 /* Initializes 'oc' appropriately as a management service controller for
3370 * The caller must free oc->target when it is no longer needed. */
3372 bridge_ofproto_controller_for_mgmt(const struct bridge *br,
3373 struct ofproto_controller *oc)
3375 oc->target = xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br->name);
3376 oc->max_backoff = 0;
3377 oc->probe_interval = 60;
3378 oc->band = OFPROTO_OUT_OF_BAND;
3380 oc->burst_limit = 0;
3381 oc->enable_async_msgs = true;
3385 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
3387 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller *c,
3388 struct ofproto_controller *oc)
3392 oc->target = c->target;
3393 oc->max_backoff = c->max_backoff ? *c->max_backoff / 1000 : 8;
3394 oc->probe_interval = c->inactivity_probe ? *c->inactivity_probe / 1000 : 5;
3395 oc->band = (!c->connection_mode || !strcmp(c->connection_mode, "in-band")
3396 ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND);
3397 oc->rate_limit = c->controller_rate_limit ? *c->controller_rate_limit : 0;
3398 oc->burst_limit = (c->controller_burst_limit
3399 ? *c->controller_burst_limit : 0);
3400 oc->enable_async_msgs = (!c->enable_async_messages
3401 || *c->enable_async_messages);
3402 dscp = smap_get_int(&c->other_config, "dscp", DSCP_DEFAULT);
3403 if (dscp < 0 || dscp > 63) {
3404 dscp = DSCP_DEFAULT;
3409 /* Configures the IP stack for 'br''s local interface properly according to the
3410 * configuration in 'c'. */
3412 bridge_configure_local_iface_netdev(struct bridge *br,
3413 struct ovsrec_controller *c)
3415 struct netdev *netdev;
3416 struct in_addr mask, gateway;
3418 struct iface *local_iface;
3421 /* If there's no local interface or no IP address, give up. */
3422 local_iface = iface_from_ofp_port(br, OFPP_LOCAL);
3423 if (!local_iface || !c->local_ip
3424 || !inet_pton(AF_INET, c->local_ip, &ip)) {
3428 /* Bring up the local interface. */
3429 netdev = local_iface->netdev;
3430 netdev_turn_flags_on(netdev, NETDEV_UP, NULL);
3432 /* Configure the IP address and netmask. */
3433 if (!c->local_netmask
3434 || !inet_pton(AF_INET, c->local_netmask, &mask)
3436 mask.s_addr = guess_netmask(ip.s_addr);
3438 if (!netdev_set_in4(netdev, ip, mask)) {
3439 VLOG_INFO("bridge %s: configured IP address "IP_FMT", netmask "IP_FMT,
3440 br->name, IP_ARGS(ip.s_addr), IP_ARGS(mask.s_addr));
3443 /* Configure the default gateway. */
3444 if (c->local_gateway
3445 && inet_pton(AF_INET, c->local_gateway, &gateway)
3446 && gateway.s_addr) {
3447 if (!netdev_add_router(netdev, gateway)) {
3448 VLOG_INFO("bridge %s: configured gateway "IP_FMT,
3449 br->name, IP_ARGS(gateway.s_addr));
3454 /* Returns true if 'a' and 'b' are the same except that any number of slashes
3455 * in either string are treated as equal to any number of slashes in the other,
3456 * e.g. "x///y" is equal to "x/y".
3458 * Also, if 'b_stoplen' bytes from 'b' are found to be equal to corresponding
3459 * bytes from 'a', the function considers this success. Specify 'b_stoplen' as
3460 * SIZE_MAX to compare all of 'a' to all of 'b' rather than just a prefix of
3461 * 'b' against a prefix of 'a'.
3464 equal_pathnames(const char *a, const char *b, size_t b_stoplen)
3466 const char *b_start = b;
3468 if (b - b_start >= b_stoplen) {
3470 } else if (*a != *b) {
3472 } else if (*a == '/') {
3473 a += strspn(a, "/");
3474 b += strspn(b, "/");
3475 } else if (*a == '\0') {
3485 bridge_configure_remotes(struct bridge *br,
3486 const struct sockaddr_in *managers, size_t n_managers)
3488 bool disable_in_band;
3490 struct ovsrec_controller **controllers;
3491 size_t n_controllers;
3493 enum ofproto_fail_mode fail_mode;
3495 struct ofproto_controller *ocs;
3499 /* Check if we should disable in-band control on this bridge. */
3500 disable_in_band = smap_get_bool(&br->cfg->other_config, "disable-in-band",
3503 /* Set OpenFlow queue ID for in-band control. */
3504 ofproto_set_in_band_queue(br->ofproto,
3505 smap_get_int(&br->cfg->other_config,
3506 "in-band-queue", -1));
3508 if (disable_in_band) {
3509 ofproto_set_extra_in_band_remotes(br->ofproto, NULL, 0);
3511 ofproto_set_extra_in_band_remotes(br->ofproto, managers, n_managers);
3514 n_controllers = bridge_get_controllers(br, &controllers);
3516 ocs = xmalloc((n_controllers + 1) * sizeof *ocs);
3519 bridge_ofproto_controller_for_mgmt(br, &ocs[n_ocs++]);
3520 for (i = 0; i < n_controllers; i++) {
3521 struct ovsrec_controller *c = controllers[i];
3523 if (!strncmp(c->target, "punix:", 6)
3524 || !strncmp(c->target, "unix:", 5)) {
3525 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3528 if (!strncmp(c->target, "unix:", 5)) {
3529 /* Connect to a listening socket */
3530 whitelist = xasprintf("unix:%s/", ovs_rundir());
3531 if (strchr(c->target, '/') &&
3532 !equal_pathnames(c->target, whitelist,
3533 strlen(whitelist))) {
3534 /* Absolute path specified, but not in ovs_rundir */
3535 VLOG_ERR_RL(&rl, "bridge %s: Not connecting to socket "
3536 "controller \"%s\" due to possibility for "
3537 "remote exploit. Instead, specify socket "
3538 "in whitelisted \"%s\" or connect to "
3539 "\"unix:%s/%s.mgmt\" (which is always "
3540 "available without special configuration).",
3541 br->name, c->target, whitelist,
3542 ovs_rundir(), br->name);
3547 whitelist = xasprintf("punix:%s/%s.controller",
3548 ovs_rundir(), br->name);
3549 if (!equal_pathnames(c->target, whitelist, SIZE_MAX)) {
3550 /* Prevent remote ovsdb-server users from accessing
3551 * arbitrary Unix domain sockets and overwriting arbitrary
3553 VLOG_ERR_RL(&rl, "bridge %s: Not adding Unix domain socket "
3554 "controller \"%s\" due to possibility of "
3555 "overwriting local files. Instead, specify "
3556 "whitelisted \"%s\" or connect to "
3557 "\"unix:%s/%s.mgmt\" (which is always "
3558 "available without special configuration).",
3559 br->name, c->target, whitelist,
3560 ovs_rundir(), br->name);
3569 bridge_configure_local_iface_netdev(br, c);
3570 bridge_ofproto_controller_from_ovsrec(c, &ocs[n_ocs]);
3571 if (disable_in_band) {
3572 ocs[n_ocs].band = OFPROTO_OUT_OF_BAND;
3577 ofproto_set_controllers(br->ofproto, ocs, n_ocs,
3578 bridge_get_allowed_versions(br));
3579 free(ocs[0].target); /* From bridge_ofproto_controller_for_mgmt(). */
3582 /* Set the fail-mode. */
3583 fail_mode = !br->cfg->fail_mode
3584 || !strcmp(br->cfg->fail_mode, "standalone")
3585 ? OFPROTO_FAIL_STANDALONE
3586 : OFPROTO_FAIL_SECURE;
3587 ofproto_set_fail_mode(br->ofproto, fail_mode);
3589 /* Configure OpenFlow controller connection snooping. */
3590 if (!ofproto_has_snoops(br->ofproto)) {
3594 sset_add_and_free(&snoops, xasprintf("punix:%s/%s.snoop",
3595 ovs_rundir(), br->name));
3596 ofproto_set_snoops(br->ofproto, &snoops);
3597 sset_destroy(&snoops);
3602 bridge_configure_tables(struct bridge *br)
3604 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
3608 n_tables = ofproto_get_n_tables(br->ofproto);
3610 for (i = 0; i < n_tables; i++) {
3611 struct ofproto_table_settings s;
3612 bool use_default_prefixes = true;
3615 s.max_flows = UINT_MAX;
3618 s.n_prefix_fields = 0;
3619 memset(s.prefix_fields, ~0, sizeof(s.prefix_fields));
3621 if (j < br->cfg->n_flow_tables && i == br->cfg->key_flow_tables[j]) {
3622 struct ovsrec_flow_table *cfg = br->cfg->value_flow_tables[j++];
3625 if (cfg->n_flow_limit && *cfg->flow_limit < UINT_MAX) {
3626 s.max_flows = *cfg->flow_limit;
3628 if (cfg->overflow_policy
3629 && !strcmp(cfg->overflow_policy, "evict")) {
3631 s.groups = xmalloc(cfg->n_groups * sizeof *s.groups);
3632 for (k = 0; k < cfg->n_groups; k++) {
3633 const char *string = cfg->groups[k];
3636 msg = mf_parse_subfield__(&s.groups[k], &string);
3638 VLOG_WARN_RL(&rl, "bridge %s table %d: error parsing "
3639 "'groups' (%s)", br->name, i, msg);
3641 } else if (*string) {
3642 VLOG_WARN_RL(&rl, "bridge %s table %d: 'groups' "
3643 "element '%s' contains trailing garbage",
3644 br->name, i, cfg->groups[k]);
3650 /* Prefix lookup fields. */
3651 s.n_prefix_fields = 0;
3652 for (k = 0; k < cfg->n_prefixes; k++) {
3653 const char *name = cfg->prefixes[k];
3654 const struct mf_field *mf;
3656 if (strcmp(name, "none") == 0) {
3657 use_default_prefixes = false;
3658 s.n_prefix_fields = 0;
3661 mf = mf_from_name(name);
3663 VLOG_WARN("bridge %s: 'prefixes' with unknown field: %s",
3667 if (mf->flow_be32ofs < 0 || mf->n_bits % 32) {
3668 VLOG_WARN("bridge %s: 'prefixes' with incompatible field: "
3669 "%s", br->name, name);
3672 if (s.n_prefix_fields >= ARRAY_SIZE(s.prefix_fields)) {
3673 VLOG_WARN("bridge %s: 'prefixes' with too many fields, "
3674 "field not used: %s", br->name, name);
3677 use_default_prefixes = false;
3678 s.prefix_fields[s.n_prefix_fields++] = mf->id;
3681 if (use_default_prefixes) {
3682 /* Use default values. */
3683 s.n_prefix_fields = ARRAY_SIZE(default_prefix_fields);
3684 memcpy(s.prefix_fields, default_prefix_fields,
3685 sizeof default_prefix_fields);
3688 struct ds ds = DS_EMPTY_INITIALIZER;
3689 for (k = 0; k < s.n_prefix_fields; k++) {
3691 ds_put_char(&ds, ',');
3693 ds_put_cstr(&ds, mf_from_id(s.prefix_fields[k])->name);
3695 if (s.n_prefix_fields == 0) {
3696 ds_put_cstr(&ds, "none");
3698 VLOG_INFO("bridge %s table %d: Prefix lookup with: %s.",
3699 br->name, i, ds_cstr(&ds));
3703 ofproto_configure_table(br->ofproto, i, &s);
3707 for (; j < br->cfg->n_flow_tables; j++) {
3708 VLOG_WARN_RL(&rl, "bridge %s: ignoring configuration for flow table "
3709 "%"PRId64" not supported by this datapath", br->name,
3710 br->cfg->key_flow_tables[j]);
3715 bridge_configure_dp_desc(struct bridge *br)
3717 ofproto_set_dp_desc(br->ofproto,
3718 smap_get(&br->cfg->other_config, "dp-desc"));
3721 static struct aa_mapping *
3722 bridge_aa_mapping_find(struct bridge *br, const int64_t isid)
3724 struct aa_mapping *m;
3726 HMAP_FOR_EACH_IN_BUCKET (m,
3728 hash_bytes(&isid, sizeof isid, 0),
3730 if (isid == m->isid) {
3737 static struct aa_mapping *
3738 bridge_aa_mapping_create(struct bridge *br,
3742 struct aa_mapping *m;
3744 m = xzalloc(sizeof *m);
3748 m->br_name = xstrdup(br->name);
3749 hmap_insert(&br->mappings,
3751 hash_bytes(&isid, sizeof isid, 0));
3757 bridge_aa_mapping_destroy(struct aa_mapping *m)
3760 struct bridge *br = m->bridge;
3763 ofproto_aa_mapping_unregister(br->ofproto, m);
3766 hmap_remove(&br->mappings, &m->hmap_node);
3775 bridge_aa_mapping_configure(struct aa_mapping *m)
3777 struct aa_mapping_settings s;
3783 ofproto_aa_mapping_register(m->bridge->ofproto, m, &s);
3789 bridge_configure_aa(struct bridge *br)
3791 const struct ovsdb_datum *mc;
3792 struct ovsrec_autoattach *auto_attach = br->cfg->auto_attach;
3793 struct aa_settings aa_s;
3794 struct aa_mapping *m, *next;
3798 ofproto_set_aa(br->ofproto, NULL, NULL);
3802 memset(&aa_s, 0, sizeof aa_s);
3803 aa_s.system_description = auto_attach->system_description;
3804 aa_s.system_name = auto_attach->system_name;
3805 ofproto_set_aa(br->ofproto, NULL, &aa_s);
3807 mc = ovsrec_autoattach_get_mappings(auto_attach,
3809 OVSDB_TYPE_INTEGER);
3810 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mappings) {
3811 union ovsdb_atom atom;
3813 atom.integer = m->isid;
3814 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
3815 VLOG_INFO("Deleting isid=%"PRIu32", vlan=%"PRIu16,
3817 bridge_aa_mapping_destroy(m);
3821 /* Add new mappings and reconfigure existing ones. */
3822 for (i = 0; i < auto_attach->n_mappings; ++i) {
3823 struct aa_mapping *m =
3824 bridge_aa_mapping_find(br, auto_attach->key_mappings[i]);
3827 VLOG_INFO("Adding isid=%"PRId64", vlan=%"PRId64,
3828 auto_attach->key_mappings[i],
3829 auto_attach->value_mappings[i]);
3830 m = bridge_aa_mapping_create(br,
3831 auto_attach->key_mappings[i],
3832 auto_attach->value_mappings[i]);
3834 if (!bridge_aa_mapping_configure(m)) {
3835 bridge_aa_mapping_destroy(m);
3842 bridge_aa_need_refresh(struct bridge *br)
3844 return ofproto_aa_vlan_get_queue_size(br->ofproto) > 0;
3848 bridge_aa_update_trunks(struct port *port, struct bridge_aa_vlan *m)
3850 int64_t *trunks = NULL;
3852 bool found = false, reconfigure = false;
3854 for (i = 0; i < port->cfg->n_trunks; i++) {
3855 if (port->cfg->trunks[i] == m->vlan) {
3862 case BRIDGE_AA_VLAN_OPER_ADD:
3864 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks + 1));
3866 for (i = 0; i < port->cfg->n_trunks; i++) {
3867 trunks[i] = port->cfg->trunks[i];
3869 trunks[i++] = m->vlan;
3875 case BRIDGE_AA_VLAN_OPER_REMOVE:
3879 trunks = xmalloc(sizeof *trunks * (port->cfg->n_trunks - 1));
3881 for (i = 0; i < port->cfg->n_trunks; i++) {
3882 if (port->cfg->trunks[i] != m->vlan) {
3883 trunks[j++] = port->cfg->trunks[i];
3892 case BRIDGE_AA_VLAN_OPER_UNDEF:
3894 VLOG_WARN("unrecognized operation %u", m->oper);
3899 /* VLAN switching under trunk mode cause the trunk port to switch all
3900 * VLANs, see ovs-vswitchd.conf.db
3903 static char *vlan_mode_access = "access";
3904 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_access);
3908 static char *vlan_mode_trunk = "trunk";
3909 ovsrec_port_set_vlan_mode(port->cfg, vlan_mode_trunk);
3912 ovsrec_port_set_trunks(port->cfg, trunks, i);
3914 /* Force reconfigure of the port. */
3915 port_configure(port);
3920 bridge_aa_refresh_queued(struct bridge *br)
3922 struct ovs_list *list = xmalloc(sizeof *list);
3923 struct bridge_aa_vlan *node;
3926 ofproto_aa_vlan_get_queued(br->ofproto, list);
3928 LIST_FOR_EACH(node, list_node, list) {
3931 VLOG_INFO("ifname=%s, vlan=%u, oper=%u", node->port_name, node->vlan,
3934 port = port_lookup(br, node->port_name);
3936 bridge_aa_update_trunks(port, node);
3939 list_remove(&node->list_node);
3940 free(node->port_name);
3948 /* Port functions. */
3950 static struct port *
3951 port_create(struct bridge *br, const struct ovsrec_port *cfg)
3955 port = xzalloc(sizeof *port);
3957 port->name = xstrdup(cfg->name);
3959 list_init(&port->ifaces);
3961 hmap_insert(&br->ports, &port->hmap_node, hash_string(port->name, 0));
3965 /* Deletes interfaces from 'port' that are no longer configured for it. */
3967 port_del_ifaces(struct port *port)
3969 struct iface *iface, *next;
3970 struct sset new_ifaces;
3973 /* Collect list of new interfaces. */
3974 sset_init(&new_ifaces);
3975 for (i = 0; i < port->cfg->n_interfaces; i++) {
3976 const char *name = port->cfg->interfaces[i]->name;
3977 const char *type = port->cfg->interfaces[i]->type;
3978 if (strcmp(type, "null")) {
3979 sset_add(&new_ifaces, name);
3983 /* Get rid of deleted interfaces. */
3984 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
3985 if (!sset_contains(&new_ifaces, iface->name)) {
3986 iface_destroy(iface);
3990 sset_destroy(&new_ifaces);
3994 port_destroy(struct port *port)
3997 struct bridge *br = port->bridge;
3998 struct iface *iface, *next;
4001 ofproto_bundle_unregister(br->ofproto, port);
4004 LIST_FOR_EACH_SAFE (iface, next, port_elem, &port->ifaces) {
4005 iface_destroy__(iface);
4008 hmap_remove(&br->ports, &port->hmap_node);
4014 static struct port *
4015 port_lookup(const struct bridge *br, const char *name)
4019 HMAP_FOR_EACH_WITH_HASH (port, hmap_node, hash_string(name, 0),
4021 if (!strcmp(port->name, name)) {
4029 enable_lacp(struct port *port, bool *activep)
4031 if (!port->cfg->lacp) {
4032 /* XXX when LACP implementation has been sufficiently tested, enable by
4033 * default and make active on bonded ports. */
4035 } else if (!strcmp(port->cfg->lacp, "off")) {
4037 } else if (!strcmp(port->cfg->lacp, "active")) {
4040 } else if (!strcmp(port->cfg->lacp, "passive")) {
4044 VLOG_WARN("port %s: unknown LACP mode %s",
4045 port->name, port->cfg->lacp);
4050 static struct lacp_settings *
4051 port_configure_lacp(struct port *port, struct lacp_settings *s)
4053 const char *lacp_time, *system_id;
4056 if (!enable_lacp(port, &s->active)) {
4060 s->name = port->name;
4062 system_id = smap_get(&port->cfg->other_config, "lacp-system-id");
4064 if (!ovs_scan(system_id, ETH_ADDR_SCAN_FMT,
4065 ETH_ADDR_SCAN_ARGS(s->id))) {
4066 VLOG_WARN("port %s: LACP system ID (%s) must be an Ethernet"
4067 " address.", port->name, system_id);
4071 memcpy(s->id, port->bridge->ea, ETH_ADDR_LEN);
4074 if (eth_addr_is_zero(s->id)) {
4075 VLOG_WARN("port %s: Invalid zero LACP system ID.", port->name);
4079 /* Prefer bondable links if unspecified. */
4080 priority = smap_get_int(&port->cfg->other_config, "lacp-system-priority",
4082 s->priority = (priority > 0 && priority <= UINT16_MAX
4084 : UINT16_MAX - !list_is_short(&port->ifaces));
4086 lacp_time = smap_get(&port->cfg->other_config, "lacp-time");
4087 s->fast = lacp_time && !strcasecmp(lacp_time, "fast");
4089 s->fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4090 "lacp-fallback-ab", false);
4096 iface_configure_lacp(struct iface *iface, struct lacp_slave_settings *s)
4098 int priority, portid, key;
4100 portid = smap_get_int(&iface->cfg->other_config, "lacp-port-id", 0);
4101 priority = smap_get_int(&iface->cfg->other_config, "lacp-port-priority",
4103 key = smap_get_int(&iface->cfg->other_config, "lacp-aggregation-key", 0);
4105 if (portid <= 0 || portid > UINT16_MAX) {
4106 portid = ofp_to_u16(iface->ofp_port);
4109 if (priority <= 0 || priority > UINT16_MAX) {
4110 priority = UINT16_MAX;
4113 if (key < 0 || key > UINT16_MAX) {
4117 s->name = iface->name;
4119 s->priority = priority;
4124 port_configure_bond(struct port *port, struct bond_settings *s)
4126 const char *detect_s;
4127 struct iface *iface;
4129 int miimon_interval;
4131 s->name = port->name;
4133 if (port->cfg->bond_mode) {
4134 if (!bond_mode_from_string(&s->balance, port->cfg->bond_mode)) {
4135 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
4136 port->name, port->cfg->bond_mode,
4137 bond_mode_to_string(s->balance));
4140 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
4142 /* XXX: Post version 1.5.*, the default bond_mode changed from SLB to
4143 * active-backup. At some point we should remove this warning. */
4144 VLOG_WARN_RL(&rl, "port %s: Using the default bond_mode %s. Note that"
4145 " in previous versions, the default bond_mode was"
4146 " balance-slb", port->name,
4147 bond_mode_to_string(s->balance));
4149 if (s->balance == BM_SLB && port->bridge->cfg->n_flood_vlans) {
4150 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
4151 "please use another bond type or disable flood_vlans",
4155 miimon_interval = smap_get_int(&port->cfg->other_config,
4156 "bond-miimon-interval", 0);
4157 if (miimon_interval <= 0) {
4158 miimon_interval = 200;
4161 detect_s = smap_get(&port->cfg->other_config, "bond-detect-mode");
4162 if (!detect_s || !strcmp(detect_s, "carrier")) {
4163 miimon_interval = 0;
4164 } else if (strcmp(detect_s, "miimon")) {
4165 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
4166 "defaulting to carrier", port->name, detect_s);
4167 miimon_interval = 0;
4170 s->up_delay = MAX(0, port->cfg->bond_updelay);
4171 s->down_delay = MAX(0, port->cfg->bond_downdelay);
4172 s->basis = smap_get_int(&port->cfg->other_config, "bond-hash-basis", 0);
4173 s->rebalance_interval = smap_get_int(&port->cfg->other_config,
4174 "bond-rebalance-interval", 10000);
4175 if (s->rebalance_interval && s->rebalance_interval < 1000) {
4176 s->rebalance_interval = 1000;
4179 s->lacp_fallback_ab_cfg = smap_get_bool(&port->cfg->other_config,
4180 "lacp-fallback-ab", false);
4182 LIST_FOR_EACH (iface, port_elem, &port->ifaces) {
4183 netdev_set_miimon_interval(iface->netdev, miimon_interval);
4186 mac_s = port->cfg->bond_active_slave;
4187 if (!mac_s || !ovs_scan(mac_s, ETH_ADDR_SCAN_FMT,
4188 ETH_ADDR_SCAN_ARGS(s->active_slave_mac))) {
4189 /* OVSDB did not store the last active interface */
4190 memset(s->active_slave_mac, 0, sizeof(s->active_slave_mac));
4194 /* Returns true if 'port' is synthetic, that is, if we constructed it locally
4195 * instead of obtaining it from the database. */
4197 port_is_synthetic(const struct port *port)
4199 return ovsdb_idl_row_is_synthetic(&port->cfg->header_);
4202 /* Interface functions. */
4205 iface_is_internal(const struct ovsrec_interface *iface,
4206 const struct ovsrec_bridge *br)
4208 /* The local port and "internal" ports are always "internal". */
4209 return !strcmp(iface->type, "internal") || !strcmp(iface->name, br->name);
4212 /* Returns the correct network device type for interface 'iface' in bridge
4215 iface_get_type(const struct ovsrec_interface *iface,
4216 const struct ovsrec_bridge *br)
4220 /* The local port always has type "internal". Other ports take
4221 * their type from the database and default to "system" if none is
4223 if (iface_is_internal(iface, br)) {
4226 type = iface->type[0] ? iface->type : "system";
4229 return ofproto_port_open_type(br->datapath_type, type);
4233 iface_destroy__(struct iface *iface)
4236 struct port *port = iface->port;
4237 struct bridge *br = port->bridge;
4239 if (br->ofproto && iface->ofp_port != OFPP_NONE) {
4240 ofproto_port_unregister(br->ofproto, iface->ofp_port);
4243 if (iface->ofp_port != OFPP_NONE) {
4244 hmap_remove(&br->ifaces, &iface->ofp_port_node);
4247 list_remove(&iface->port_elem);
4248 hmap_remove(&br->iface_by_name, &iface->name_node);
4250 /* The user is changing configuration here, so netdev_remove needs to be
4251 * used as opposed to netdev_close */
4252 netdev_remove(iface->netdev);
4260 iface_destroy(struct iface *iface)
4263 struct port *port = iface->port;
4265 iface_destroy__(iface);
4266 if (list_is_empty(&port->ifaces)) {
4272 static struct iface *
4273 iface_lookup(const struct bridge *br, const char *name)
4275 struct iface *iface;
4277 HMAP_FOR_EACH_WITH_HASH (iface, name_node, hash_string(name, 0),
4278 &br->iface_by_name) {
4279 if (!strcmp(iface->name, name)) {
4287 static struct iface *
4288 iface_find(const char *name)
4290 const struct bridge *br;
4292 HMAP_FOR_EACH (br, node, &all_bridges) {
4293 struct iface *iface = iface_lookup(br, name);
4302 static struct iface *
4303 iface_from_ofp_port(const struct bridge *br, ofp_port_t ofp_port)
4305 struct iface *iface;
4307 HMAP_FOR_EACH_IN_BUCKET (iface, ofp_port_node, hash_ofp_port(ofp_port),
4309 if (iface->ofp_port == ofp_port) {
4316 /* Set Ethernet address of 'iface', if one is specified in the configuration
4319 iface_set_mac(const struct bridge *br, const struct port *port, struct iface *iface)
4321 uint8_t ea[ETH_ADDR_LEN], *mac = NULL;
4322 struct iface *hw_addr_iface;
4324 if (strcmp(iface->type, "internal")) {
4328 if (iface->cfg->mac && eth_addr_from_string(iface->cfg->mac, ea)) {
4330 } else if (port->cfg->fake_bridge) {
4331 /* Fake bridge and no MAC set in the configuration. Pick a local one. */
4332 find_local_hw_addr(br, ea, port, &hw_addr_iface);
4337 if (iface->ofp_port == OFPP_LOCAL) {
4338 VLOG_ERR("interface %s: ignoring mac in Interface record "
4339 "(use Bridge record to set local port's mac)",
4341 } else if (eth_addr_is_multicast(mac)) {
4342 VLOG_ERR("interface %s: cannot set MAC to multicast address",
4345 int error = netdev_set_etheraddr(iface->netdev, mac);
4347 VLOG_ERR("interface %s: setting MAC failed (%s)",
4348 iface->name, ovs_strerror(error));
4354 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
4356 iface_set_ofport(const struct ovsrec_interface *if_cfg, ofp_port_t ofport)
4358 if (if_cfg && !ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4359 int64_t port = ofport == OFPP_NONE ? -1 : ofp_to_u16(ofport);
4360 ovsrec_interface_set_ofport(if_cfg, &port, 1);
4364 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
4365 * sets the "ofport" field to -1.
4367 * This is appropriate when 'if_cfg''s interface cannot be created or is
4368 * otherwise invalid. */
4370 iface_clear_db_record(const struct ovsrec_interface *if_cfg, char *errp)
4372 if (!ovsdb_idl_row_is_synthetic(&if_cfg->header_)) {
4373 iface_set_ofport(if_cfg, OFPP_NONE);
4374 ovsrec_interface_set_error(if_cfg, errp);
4375 ovsrec_interface_set_status(if_cfg, NULL);
4376 ovsrec_interface_set_admin_state(if_cfg, NULL);
4377 ovsrec_interface_set_duplex(if_cfg, NULL);
4378 ovsrec_interface_set_link_speed(if_cfg, NULL, 0);
4379 ovsrec_interface_set_link_state(if_cfg, NULL);
4380 ovsrec_interface_set_mac_in_use(if_cfg, NULL);
4381 ovsrec_interface_set_mtu(if_cfg, NULL, 0);
4382 ovsrec_interface_set_cfm_fault(if_cfg, NULL, 0);
4383 ovsrec_interface_set_cfm_fault_status(if_cfg, NULL, 0);
4384 ovsrec_interface_set_cfm_remote_mpids(if_cfg, NULL, 0);
4385 ovsrec_interface_set_lacp_current(if_cfg, NULL, 0);
4386 ovsrec_interface_set_statistics(if_cfg, NULL, NULL, 0);
4387 ovsrec_interface_set_ifindex(if_cfg, NULL, 0);
4392 queue_ids_include(const struct ovsdb_datum *queues, int64_t target)
4394 union ovsdb_atom atom;
4396 atom.integer = target;
4397 return ovsdb_datum_find_key(queues, &atom, OVSDB_TYPE_INTEGER) != UINT_MAX;
4401 iface_configure_qos(struct iface *iface, const struct ovsrec_qos *qos)
4403 struct ofpbuf queues_buf;
4405 ofpbuf_init(&queues_buf, 0);
4407 if (!qos || qos->type[0] == '\0') {
4408 netdev_set_qos(iface->netdev, NULL, NULL);
4410 const struct ovsdb_datum *queues;
4411 struct netdev_queue_dump dump;
4412 unsigned int queue_id;
4413 struct smap details;
4417 /* Configure top-level Qos for 'iface'. */
4418 netdev_set_qos(iface->netdev, qos->type, &qos->other_config);
4420 /* Deconfigure queues that were deleted. */
4421 queues = ovsrec_qos_get_queues(qos, OVSDB_TYPE_INTEGER,
4423 smap_init(&details);
4424 NETDEV_QUEUE_FOR_EACH (&queue_id, &details, &dump, iface->netdev) {
4425 if (!queue_ids_include(queues, queue_id)) {
4426 netdev_delete_queue(iface->netdev, queue_id);
4429 smap_destroy(&details);
4431 /* Configure queues for 'iface'. */
4433 for (i = 0; i < qos->n_queues; i++) {
4434 const struct ovsrec_queue *queue = qos->value_queues[i];
4435 unsigned int queue_id = qos->key_queues[i];
4437 if (queue_id == 0) {
4441 if (queue->n_dscp == 1) {
4442 struct ofproto_port_queue *port_queue;
4444 port_queue = ofpbuf_put_uninit(&queues_buf,
4445 sizeof *port_queue);
4446 port_queue->queue = queue_id;
4447 port_queue->dscp = queue->dscp[0];
4450 netdev_set_queue(iface->netdev, queue_id, &queue->other_config);
4453 struct smap details;
4455 smap_init(&details);
4456 netdev_set_queue(iface->netdev, 0, &details);
4457 smap_destroy(&details);
4461 if (iface->ofp_port != OFPP_NONE) {
4462 const struct ofproto_port_queue *port_queues = queues_buf.data;
4463 size_t n_queues = queues_buf.size / sizeof *port_queues;
4465 ofproto_port_set_queues(iface->port->bridge->ofproto, iface->ofp_port,
4466 port_queues, n_queues);
4469 netdev_set_policing(iface->netdev,
4470 MIN(UINT32_MAX, iface->cfg->ingress_policing_rate),
4471 MIN(UINT32_MAX, iface->cfg->ingress_policing_burst));
4473 ofpbuf_uninit(&queues_buf);
4477 iface_configure_cfm(struct iface *iface)
4479 const struct ovsrec_interface *cfg = iface->cfg;
4480 const char *opstate_str;
4481 const char *cfm_ccm_vlan;
4482 struct cfm_settings s;
4483 struct smap netdev_args;
4485 if (!cfg->n_cfm_mpid) {
4486 ofproto_port_clear_cfm(iface->port->bridge->ofproto, iface->ofp_port);
4490 s.check_tnl_key = false;
4491 smap_init(&netdev_args);
4492 if (!netdev_get_config(iface->netdev, &netdev_args)) {
4493 const char *key = smap_get(&netdev_args, "key");
4494 const char *in_key = smap_get(&netdev_args, "in_key");
4496 s.check_tnl_key = (key && !strcmp(key, "flow"))
4497 || (in_key && !strcmp(in_key, "flow"));
4499 smap_destroy(&netdev_args);
4501 s.mpid = *cfg->cfm_mpid;
4502 s.interval = smap_get_int(&iface->cfg->other_config, "cfm_interval", 0);
4503 cfm_ccm_vlan = smap_get(&iface->cfg->other_config, "cfm_ccm_vlan");
4504 s.ccm_pcp = smap_get_int(&iface->cfg->other_config, "cfm_ccm_pcp", 0);
4506 if (s.interval <= 0) {
4510 if (!cfm_ccm_vlan) {
4512 } else if (!strcasecmp("random", cfm_ccm_vlan)) {
4513 s.ccm_vlan = CFM_RANDOM_VLAN;
4515 s.ccm_vlan = atoi(cfm_ccm_vlan);
4516 if (s.ccm_vlan == CFM_RANDOM_VLAN) {
4521 s.extended = smap_get_bool(&iface->cfg->other_config, "cfm_extended",
4523 s.demand = smap_get_bool(&iface->cfg->other_config, "cfm_demand", false);
4525 opstate_str = smap_get(&iface->cfg->other_config, "cfm_opstate");
4526 s.opup = !opstate_str || !strcasecmp("up", opstate_str);
4528 ofproto_port_set_cfm(iface->port->bridge->ofproto, iface->ofp_port, &s);
4531 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
4532 * instead of obtaining it from the database. */
4534 iface_is_synthetic(const struct iface *iface)
4536 return ovsdb_idl_row_is_synthetic(&iface->cfg->header_);
4540 iface_validate_ofport__(size_t n, int64_t *ofport)
4542 return (n && *ofport >= 1 && *ofport < ofp_to_u16(OFPP_MAX)
4543 ? u16_to_ofp(*ofport)
4548 iface_get_requested_ofp_port(const struct ovsrec_interface *cfg)
4550 return iface_validate_ofport__(cfg->n_ofport_request, cfg->ofport_request);
4554 iface_pick_ofport(const struct ovsrec_interface *cfg)
4556 ofp_port_t requested_ofport = iface_get_requested_ofp_port(cfg);
4557 return (requested_ofport != OFPP_NONE
4559 : iface_validate_ofport__(cfg->n_ofport, cfg->ofport));
4562 /* Port mirroring. */
4564 static struct mirror *
4565 mirror_find_by_uuid(struct bridge *br, const struct uuid *uuid)
4569 HMAP_FOR_EACH_IN_BUCKET (m, hmap_node, uuid_hash(uuid), &br->mirrors) {
4570 if (uuid_equals(uuid, &m->uuid)) {
4578 bridge_configure_mirrors(struct bridge *br)
4580 const struct ovsdb_datum *mc;
4581 unsigned long *flood_vlans;
4582 struct mirror *m, *next;
4585 /* Get rid of deleted mirrors. */
4586 mc = ovsrec_bridge_get_mirrors(br->cfg, OVSDB_TYPE_UUID);
4587 HMAP_FOR_EACH_SAFE (m, next, hmap_node, &br->mirrors) {
4588 union ovsdb_atom atom;
4590 atom.uuid = m->uuid;
4591 if (ovsdb_datum_find_key(mc, &atom, OVSDB_TYPE_UUID) == UINT_MAX) {
4596 /* Add new mirrors and reconfigure existing ones. */
4597 for (i = 0; i < br->cfg->n_mirrors; i++) {
4598 const struct ovsrec_mirror *cfg = br->cfg->mirrors[i];
4599 struct mirror *m = mirror_find_by_uuid(br, &cfg->header_.uuid);
4601 m = mirror_create(br, cfg);
4604 if (!mirror_configure(m)) {
4609 /* Update flooded vlans (for RSPAN). */
4610 flood_vlans = vlan_bitmap_from_array(br->cfg->flood_vlans,
4611 br->cfg->n_flood_vlans);
4612 ofproto_set_flood_vlans(br->ofproto, flood_vlans);
4613 bitmap_free(flood_vlans);
4616 static struct mirror *
4617 mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
4621 m = xzalloc(sizeof *m);
4622 m->uuid = cfg->header_.uuid;
4623 hmap_insert(&br->mirrors, &m->hmap_node, uuid_hash(&m->uuid));
4625 m->name = xstrdup(cfg->name);
4631 mirror_destroy(struct mirror *m)
4634 struct bridge *br = m->bridge;
4637 ofproto_mirror_unregister(br->ofproto, m);
4640 hmap_remove(&br->mirrors, &m->hmap_node);
4647 mirror_collect_ports(struct mirror *m,
4648 struct ovsrec_port **in_ports, int n_in_ports,
4649 void ***out_portsp, size_t *n_out_portsp)
4651 void **out_ports = xmalloc(n_in_ports * sizeof *out_ports);
4652 size_t n_out_ports = 0;
4655 for (i = 0; i < n_in_ports; i++) {
4656 const char *name = in_ports[i]->name;
4657 struct port *port = port_lookup(m->bridge, name);
4659 out_ports[n_out_ports++] = port;
4661 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
4662 "port %s", m->bridge->name, m->name, name);
4665 *out_portsp = out_ports;
4666 *n_out_portsp = n_out_ports;
4670 mirror_configure(struct mirror *m)
4672 const struct ovsrec_mirror *cfg = m->cfg;
4673 struct ofproto_mirror_settings s;
4676 if (strcmp(cfg->name, m->name)) {
4678 m->name = xstrdup(cfg->name);
4682 /* Get output port or VLAN. */
4683 if (cfg->output_port) {
4684 s.out_bundle = port_lookup(m->bridge, cfg->output_port->name);
4685 if (!s.out_bundle) {
4686 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
4687 m->bridge->name, m->name);
4690 s.out_vlan = UINT16_MAX;
4692 if (cfg->output_vlan) {
4693 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
4694 "output vlan; ignoring output vlan",
4695 m->bridge->name, m->name);
4697 } else if (cfg->output_vlan) {
4698 /* The database should prevent invalid VLAN values. */
4699 s.out_bundle = NULL;
4700 s.out_vlan = *cfg->output_vlan;
4702 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
4703 m->bridge->name, m->name);
4707 /* Get port selection. */
4708 if (cfg->select_all) {
4709 size_t n_ports = hmap_count(&m->bridge->ports);
4710 void **ports = xmalloc(n_ports * sizeof *ports);
4715 HMAP_FOR_EACH (port, hmap_node, &m->bridge->ports) {
4725 /* Get ports, dropping ports that don't exist.
4726 * The IDL ensures that there are no duplicates. */
4727 mirror_collect_ports(m, cfg->select_src_port, cfg->n_select_src_port,
4728 &s.srcs, &s.n_srcs);
4729 mirror_collect_ports(m, cfg->select_dst_port, cfg->n_select_dst_port,
4730 &s.dsts, &s.n_dsts);
4733 /* Get VLAN selection. */
4734 s.src_vlans = vlan_bitmap_from_array(cfg->select_vlan, cfg->n_select_vlan);
4737 ofproto_mirror_register(m->bridge->ofproto, m, &s);
4740 if (s.srcs != s.dsts) {
4749 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4751 * This is deprecated. It is only for compatibility with broken device drivers
4752 * in old versions of Linux that do not properly support VLANs when VLAN
4753 * devices are not used. When broken device drivers are no longer in
4754 * widespread use, we will delete these interfaces. */
4756 static struct ovsrec_port **recs;
4757 static size_t n_recs, allocated_recs;
4759 /* Adds 'rec' to a list of recs that have to be destroyed when the VLAN
4760 * splinters are reconfigured. */
4762 register_rec(struct ovsrec_port *rec)
4764 if (n_recs >= allocated_recs) {
4765 recs = x2nrealloc(recs, &allocated_recs, sizeof *recs);
4767 recs[n_recs++] = rec;
4770 /* Frees all of the ports registered with register_reg(). */
4772 free_registered_recs(void)
4776 for (i = 0; i < n_recs; i++) {
4777 struct ovsrec_port *port = recs[i];
4780 for (j = 0; j < port->n_interfaces; j++) {
4781 struct ovsrec_interface *iface = port->interfaces[j];
4786 smap_destroy(&port->other_config);
4787 free(port->interfaces);
4795 /* Returns true if VLAN splinters are enabled on 'iface_cfg', false
4798 vlan_splinters_is_enabled(const struct ovsrec_interface *iface_cfg)
4800 return smap_get_bool(&iface_cfg->other_config, "enable-vlan-splinters",
4804 /* Figures out the set of VLANs that are in use for the purpose of VLAN
4807 * If VLAN splinters are enabled on at least one interface and any VLANs are in
4808 * use, returns a 4096-bit bitmap with a 1-bit for each in-use VLAN (bits 0 and
4809 * 4095 will not be set). The caller is responsible for freeing the bitmap,
4812 * If VLANs splinters are not enabled on any interface or if no VLANs are in
4813 * use, returns NULL.
4815 * Updates 'vlan_splinters_enabled_anywhere'. */
4816 static unsigned long int *
4817 collect_splinter_vlans(const struct ovsrec_open_vswitch *ovs_cfg)
4819 unsigned long int *splinter_vlans;
4820 struct sset splinter_ifaces;
4821 const char *real_dev_name;
4822 struct shash *real_devs;
4823 struct shash_node *node;
4827 /* Free space allocated for synthesized ports and interfaces, since we're
4828 * in the process of reconstructing all of them. */
4829 free_registered_recs();
4831 splinter_vlans = bitmap_allocate(4096);
4832 sset_init(&splinter_ifaces);
4833 vlan_splinters_enabled_anywhere = false;
4834 for (i = 0; i < ovs_cfg->n_bridges; i++) {
4835 struct ovsrec_bridge *br_cfg = ovs_cfg->bridges[i];
4838 for (j = 0; j < br_cfg->n_ports; j++) {
4839 struct ovsrec_port *port_cfg = br_cfg->ports[j];
4842 for (k = 0; k < port_cfg->n_interfaces; k++) {
4843 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[k];
4845 if (vlan_splinters_is_enabled(iface_cfg)) {
4846 vlan_splinters_enabled_anywhere = true;
4847 sset_add(&splinter_ifaces, iface_cfg->name);
4848 vlan_bitmap_from_array__(port_cfg->trunks,
4854 if (port_cfg->tag && *port_cfg->tag > 0 && *port_cfg->tag < 4095) {
4855 bitmap_set1(splinter_vlans, *port_cfg->tag);
4860 if (!vlan_splinters_enabled_anywhere) {
4861 free(splinter_vlans);
4862 sset_destroy(&splinter_ifaces);
4866 HMAP_FOR_EACH (br, node, &all_bridges) {
4868 ofproto_get_vlan_usage(br->ofproto, splinter_vlans);
4872 /* Don't allow VLANs 0 or 4095 to be splintered. VLAN 0 should appear on
4873 * the real device. VLAN 4095 is reserved and Linux doesn't allow a VLAN
4874 * device to be created for it. */
4875 bitmap_set0(splinter_vlans, 0);
4876 bitmap_set0(splinter_vlans, 4095);
4878 /* Delete all VLAN devices that we don't need. */
4880 real_devs = vlandev_get_real_devs();
4881 SHASH_FOR_EACH (node, real_devs) {
4882 const struct vlan_real_dev *real_dev = node->data;
4883 const struct vlan_dev *vlan_dev;
4884 bool real_dev_has_splinters;
4886 real_dev_has_splinters = sset_contains(&splinter_ifaces,
4888 HMAP_FOR_EACH (vlan_dev, hmap_node, &real_dev->vlan_devs) {
4889 if (!real_dev_has_splinters
4890 || !bitmap_is_set(splinter_vlans, vlan_dev->vid)) {
4891 struct netdev *netdev;
4893 if (!netdev_open(vlan_dev->name, "system", &netdev)) {
4894 if (!netdev_get_in4(netdev, NULL, NULL) ||
4895 !netdev_get_in6(netdev, NULL)) {
4896 /* It has an IP address configured, so we don't own
4897 * it. Don't delete it. */
4899 vlandev_del(vlan_dev->name);
4901 netdev_close(netdev);
4908 /* Add all VLAN devices that we need. */
4909 SSET_FOR_EACH (real_dev_name, &splinter_ifaces) {
4912 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
4913 if (!vlandev_get_name(real_dev_name, vid)) {
4914 vlandev_add(real_dev_name, vid);
4921 sset_destroy(&splinter_ifaces);
4923 if (bitmap_scan(splinter_vlans, 1, 0, 4096) >= 4096) {
4924 free(splinter_vlans);
4927 return splinter_vlans;
4930 /* Pushes the configure of VLAN splinter port 'port' (e.g. eth0.9) down to
4933 configure_splinter_port(struct port *port)
4935 struct ofproto *ofproto = port->bridge->ofproto;
4936 ofp_port_t realdev_ofp_port;
4937 const char *realdev_name;
4938 struct iface *vlandev, *realdev;
4940 ofproto_bundle_unregister(port->bridge->ofproto, port);
4942 vlandev = CONTAINER_OF(list_front(&port->ifaces), struct iface,
4945 realdev_name = smap_get(&port->cfg->other_config, "realdev");
4946 realdev = iface_lookup(port->bridge, realdev_name);
4947 realdev_ofp_port = realdev ? realdev->ofp_port : 0;
4949 ofproto_port_set_realdev(ofproto, vlandev->ofp_port, realdev_ofp_port,
4953 static struct ovsrec_port *
4954 synthesize_splinter_port(const char *real_dev_name,
4955 const char *vlan_dev_name, int vid)
4957 struct ovsrec_interface *iface;
4958 struct ovsrec_port *port;
4960 iface = xmalloc(sizeof *iface);
4961 ovsrec_interface_init(iface);
4962 iface->name = xstrdup(vlan_dev_name);
4963 iface->type = "system";
4965 port = xmalloc(sizeof *port);
4966 ovsrec_port_init(port);
4967 port->interfaces = xmemdup(&iface, sizeof iface);
4968 port->n_interfaces = 1;
4969 port->name = xstrdup(vlan_dev_name);
4970 port->vlan_mode = "splinter";
4971 port->tag = xmalloc(sizeof *port->tag);
4974 smap_add(&port->other_config, "realdev", real_dev_name);
4980 /* For each interface with 'br' that has VLAN splinters enabled, adds a
4981 * corresponding ovsrec_port to 'ports' for each splinter VLAN marked with a
4982 * 1-bit in the 'splinter_vlans' bitmap. */
4984 add_vlan_splinter_ports(struct bridge *br,
4985 const unsigned long int *splinter_vlans,
4986 struct shash *ports)
4990 /* We iterate through 'br->cfg->ports' instead of 'ports' here because
4991 * we're modifying 'ports'. */
4992 for (i = 0; i < br->cfg->n_ports; i++) {
4993 const char *name = br->cfg->ports[i]->name;
4994 struct ovsrec_port *port_cfg = shash_find_data(ports, name);
4997 for (j = 0; j < port_cfg->n_interfaces; j++) {
4998 struct ovsrec_interface *iface_cfg = port_cfg->interfaces[j];
5000 if (vlan_splinters_is_enabled(iface_cfg)) {
5001 const char *real_dev_name;
5004 real_dev_name = iface_cfg->name;
5005 BITMAP_FOR_EACH_1 (vid, 4096, splinter_vlans) {
5006 const char *vlan_dev_name;
5008 vlan_dev_name = vlandev_get_name(real_dev_name, vid);
5010 && !shash_find(ports, vlan_dev_name)) {
5011 shash_add(ports, vlan_dev_name,
5012 synthesize_splinter_port(
5013 real_dev_name, vlan_dev_name, vid));
5022 mirror_refresh_stats(struct mirror *m)
5024 struct ofproto *ofproto = m->bridge->ofproto;
5025 uint64_t tx_packets, tx_bytes;
5026 const char *keys[2];
5028 size_t stat_cnt = 0;
5030 if (ofproto_mirror_get_stats(ofproto, m, &tx_packets, &tx_bytes)) {
5031 ovsrec_mirror_set_statistics(m->cfg, NULL, NULL, 0);
5035 if (tx_packets != UINT64_MAX) {
5036 keys[stat_cnt] = "tx_packets";
5037 values[stat_cnt] = tx_packets;
5040 if (tx_bytes != UINT64_MAX) {
5041 keys[stat_cnt] = "tx_bytes";
5042 values[stat_cnt] = tx_bytes;
5046 ovsrec_mirror_set_statistics(m->cfg, keys, values, stat_cnt);
5050 * Add registered netdev and dpif types to ovsdb to allow external
5051 * applications to query the capabilities of the Open vSwitch instance
5052 * running on the node.
5055 discover_types(const struct ovsrec_open_vswitch *cfg)
5059 /* Datapath types. */
5061 dp_enumerate_types(&types);
5062 const char **datapath_types = sset_array(&types);
5063 ovsrec_open_vswitch_set_datapath_types(cfg, datapath_types,
5064 sset_count(&types));
5065 free(datapath_types);
5066 sset_destroy(&types);
5070 netdev_enumerate_types(&types);
5071 const char **iface_types = sset_array(&types);
5072 ovsrec_open_vswitch_set_iface_types(cfg, iface_types, sset_count(&types));
5074 sset_destroy(&types);